WO2023226355A1 - Procédé et système de détection de défaillance de batterie à double ion basés sur une perception multi-source - Google Patents

Procédé et système de détection de défaillance de batterie à double ion basés sur une perception multi-source Download PDF

Info

Publication number
WO2023226355A1
WO2023226355A1 PCT/CN2022/136948 CN2022136948W WO2023226355A1 WO 2023226355 A1 WO2023226355 A1 WO 2023226355A1 CN 2022136948 W CN2022136948 W CN 2022136948W WO 2023226355 A1 WO2023226355 A1 WO 2023226355A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
battery
type
fault detection
time
Prior art date
Application number
PCT/CN2022/136948
Other languages
English (en)
Chinese (zh)
Inventor
郭媛君
安钊
杨之乐
刘祥飞
胡天宇
Original Assignee
深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Publication of WO2023226355A1 publication Critical patent/WO2023226355A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10—Energy storage using batteries

Definitions

  • the present invention relates to the technical field of battery fault diagnosis, and in particular to a bidirectional lithium-ion battery fault detection method and system based on multi-source sensing.
  • lithium-ion batteries With the development of science and technology, batteries are used more and more widely, and lithium-ion batteries have received more applications and attention due to their high operating voltage, large specific energy, and long cycle life. Furthermore, currently, bidirectional lithium batteries The research and application of ion batteries are also increasing. During actual use, lithium-ion batteries (such as bidirectional lithium-ion batteries) may malfunction, resulting in inoperability, so fault diagnosis is required.
  • the main purpose of the present invention is to provide a bidirectional lithium-ion battery fault detection method and system based on multi-source sensing, aiming to solve the problem that the inference accuracy of expert systems in the prior art is not high and the failure to integrate multiple types of information for fault detection. Problems that are not conducive to improving the accuracy of bidirectional lithium-ion battery fault detection.
  • a first aspect of the present invention provides a bidirectional lithium-ion battery fault detection method based on multi-source sensing, wherein the above-mentioned bidirectional lithium-ion battery fault detection method based on multi-source sensing includes:
  • the above-mentioned first type of data includes battery heat distribution data, charge and discharge voltage data and battery temperature change data of the above-mentioned battery to be detected.
  • the above-mentioned second type of data includes battery model data, battery capacity data, battery data of the above-mentioned battery to be detected. Usage duration data, ambient temperature data and ambient humidity data;
  • the above-mentioned target time-frequency data and the above-mentioned second type data are input into the pre-trained fault detection model, and the fault category corresponding to the above-mentioned battery to be detected at the above-mentioned current moment is obtained through the above-mentioned fault detection model.
  • the first type of data is obtained by collecting multiple data on the battery to be tested within the target time period, and the battery thermal distribution data includes multiple heat distribution maps corresponding to the battery to be tested.
  • the target time-frequency data includes a target time-frequency diagram
  • the horizontal axis of the target time-frequency diagram represents the time domain
  • the vertical axis represents the frequency domain
  • the above fault detection model is pre-trained according to the following steps:
  • Each of the above-mentioned real sample data includes first type real data, second type real data and fault label data.
  • the above-mentioned first type The type of real data includes the battery heat distribution data, charge and discharge voltage data and battery temperature change data of the above-mentioned training battery during the training time period.
  • the above-mentioned second type of real data includes the battery model data and battery data of the above-mentioned training battery during the above-mentioned training time period. Capacity data, battery life data, ambient temperature data and ambient humidity data;
  • one of the above-mentioned simulated sample data includes first type simulation data, second type simulation data and simulation tag data.
  • the above-mentioned second type simulation data is consistent with the target real sample data.
  • the second type of real data of the sample data is the same, the above-mentioned simulated label data is the same as the fault label data of the above-mentioned target real sample data, and the above-mentioned target real sample data is the real sample data corresponding to the simulated sample data;
  • the above fault detection model is trained based on the above real sample data and the above simulated sample data.
  • the above-mentioned sample simulation expansion is performed based on each of the above-mentioned real sample data to obtain multiple simulated sample data, including:
  • any real sample data obtain each type of data to be processed corresponding to the above-mentioned real sample data, perform data simulation on the above-mentioned data to be processed according to the preset data simulation steps and the statistical characteristics of the above-mentioned data to be processed, and obtain each of the above-mentioned data to be processed.
  • target simulation data corresponding to the data, and all target simulation data corresponding to one of the above-mentioned real sample data are used as the first type of simulation data corresponding to the simulation sample data corresponding to the real sample data, wherein the above-mentioned data to be processed includes the above-mentioned real sample data
  • the above preset data simulation steps include:
  • segment statistical characteristics corresponding to each of the above-mentioned time segments wherein the above-mentioned segment statistical characteristics are the statistical characteristics of the above-mentioned data to be processed in each of the above-mentioned time segments, and the above-mentioned statistical characteristics include average value, maximum value and minimum value;
  • a set of simulated data values is generated through uniform distribution in each of the above-mentioned time segments, and we obtain Simulation processing data corresponding to the above data to be processed;
  • the above-mentioned simulation processing data and the above-mentioned data to be processed are weighted and summed according to the preset weight coefficient to obtain the target simulation data corresponding to the above-mentioned data to be processed.
  • one of the above-mentioned real sample data corresponds to multiple simulated sample data.
  • the above-mentioned sample simulation expansion based on each of the above-mentioned real sample data to obtain multiple simulated sample data also includes:
  • the to-be-processed data corresponding to the above-mentioned real sample data is segmented by time length and a variety of different segmentation results are obtained.
  • a simulation corresponding to the real sample data is obtained based on each segmentation result. sample.
  • the above-mentioned preset segmentation methods include random segmentation, uniform segmentation, and segmentation based on data volatility.
  • the above-mentioned training of the above-mentioned fault detection model based on the above-mentioned real sample data and the above-mentioned simulated sample data includes:
  • the above-mentioned fault detection model is trained according to the above-mentioned training data set, and the above-mentioned fault detection model is tested according to the above-mentioned test data set until the training of the above-mentioned fault detection model is completed, wherein the above-mentioned fault detection model is a convolutional god-level network model.
  • a second aspect of the present invention provides a bidirectional lithium-ion battery fault detection system based on multi-source sensing, wherein the above-mentioned bidirectional lithium-ion battery fault detection system based on multi-source sensing includes:
  • the usage status data acquisition module is used to obtain the usage status data of the battery to be detected within a target time period, where the length of the above-mentioned target time period is a preset length, the end time of the above-mentioned target time period is the current time, and the above-mentioned usage status
  • the data includes first type data and second type data.
  • the first type data includes battery thermal distribution data, charge and discharge voltage data and battery temperature change data of the battery to be detected.
  • the second type data includes battery data of the battery to be detected. Model data, battery capacity data, battery usage time data, ambient temperature data and ambient humidity data;
  • the first data processing module is used to perform principal component analysis on the above-mentioned first type data and obtain target feature data, wherein the above-mentioned target feature data is any one of the above-mentioned first type data, or the above-mentioned target feature data is composed of Comprehensive feature data composed of multiple data combinations in the above first type of data;
  • the second data processing module is used to perform wavelet transformation on the above target characteristic data to obtain target time-frequency data
  • a fault detection module is used to input the above target time-frequency data and the above second type data into a pre-trained fault detection model, and obtain the fault category corresponding to the above-mentioned battery to be detected at the above-mentioned current moment through the above-mentioned fault detection model.
  • the usage status data of the battery to be detected within the target time period is obtained, wherein the time length of the above-mentioned target time period is the preset time period, the end time of the above-mentioned target time period is the current time, and the above-mentioned usage
  • the status data includes first type data and second type data.
  • the first type data includes battery heat distribution data, charge and discharge voltage data and battery temperature change data of the battery to be detected.
  • the second type data includes battery temperature change data of the battery to be detected.
  • a fault detection model for bidirectional lithium-ion batteries is pre-established and trained, so that battery faults can be detected through the fault detection model. Specifically, perform principal component analysis on the first type of data in the usage status data corresponding to the battery to be detected and obtain the target feature data, then perform wavelet transformation to obtain the target time-frequency data, and then combine the target time-frequency data with the second type Data is used to obtain the corresponding fault category through the fault detection model. In this way, there is no need to use an expert system for reasoning, and a variety of different usage status data can be integrated to detect possible current faults of the bidirectional lithium-ion battery, which is beneficial to improving the accuracy of bidirectional lithium-ion battery fault detection.
  • Figure 1 is a schematic flow chart of a bidirectional lithium-ion battery fault detection method based on multi-source sensing provided by an embodiment of the present invention
  • Figure 2 is a schematic structural diagram of a bidirectional lithium-ion battery fault detection system based on multi-source sensing provided by an embodiment of the present invention.
  • lithium-ion batteries With the development of science and technology, batteries are used more and more widely, and lithium-ion batteries have received more applications and attention due to their high operating voltage, large specific energy, and long cycle life. Furthermore, currently, bidirectional lithium batteries The research and application of ion batteries are also increasing. During actual use, lithium-ion batteries (such as bidirectional lithium-ion batteries) may malfunction, resulting in inoperability, so fault diagnosis is required.
  • the above-mentioned bidirectional lithium-ion battery is a dual-ion battery.
  • the anion intercalation reaction occurs in the positive electrode graphite, while the aluminum-lithium alloying reaction occurs in the aluminum negative electrode, and the discharge process is opposite. It is beneficial to increase the working voltage of the battery, and at the same time, it can reduce the quality, volume and manufacturing cost of the battery, thereby comprehensively improving the energy density of the battery and increasing the battery life.
  • bidirectional lithium-ion batteries is still under development. Therefore, there are few relevant fault sample data of bidirectional lithium-ion batteries that can be used to train the model, which is not conducive to model training.
  • the expert system is a program that uses existing knowledge to reason about actual problems. It is built on an expert database and performs reasoning and classification based on input fault information combined with corresponding computer algorithms to complete fault diagnosis and decision-making. The accuracy depends on the perfection of the expert database, which requires a large amount of engineering experience and knowledge. There are problems such as low fault tolerance rate in the expert system, low diagnostic accuracy for uncertain information, difficulty in obtaining expert knowledge and experience, and high maintenance difficulty.
  • the expert system performs fault detection, one type of data usually corresponds to one type of fault. It uses fewer data types and cannot comprehensively make judgments based on multi-source sensing data.
  • the usage status data of the battery to be detected within a target time period is obtained, wherein the time length of the above-mentioned target time period is a preset time period, and the above-mentioned target time period is The end time is the current moment.
  • the above-mentioned usage status data includes first type data and second type data.
  • the above-mentioned first type data includes battery heat distribution data, charge and discharge voltage data and battery temperature change data of the above-mentioned battery to be detected.
  • the above-mentioned second type data includes the battery model data, battery capacity data, battery usage time data, ambient temperature data and ambient humidity data of the above-mentioned battery to be detected; perform principal component analysis processing on the above-mentioned first type data and obtain target characteristic data, wherein the above-mentioned target
  • the feature data is any one of the above-mentioned first type data, or the above-mentioned target feature data is comprehensive feature data composed of multiple types of data in the above-mentioned first type data; perform wavelet transformation on the above-mentioned target feature data to obtain the target time frequency data; input the above-mentioned target time-frequency data and the above-mentioned second type data into the pre-trained fault detection model, and obtain the fault category corresponding to the above-mentioned battery to be detected at the above-mentioned current moment through the above-mentioned fault detection model.
  • a fault detection model for bidirectional lithium-ion batteries is pre-established and trained, so that battery faults can be detected through the fault detection model. Specifically, perform principal component analysis on the first type of data in the usage status data corresponding to the battery to be detected and obtain the target feature data, then perform wavelet transformation to obtain the target time-frequency data, and then combine the target time-frequency data with the second type Data is used to obtain the corresponding fault category through the fault detection model. In this way, there is no need to use an expert system for reasoning, and a variety of different usage status data can be integrated to detect possible current faults of the bidirectional lithium-ion battery, which is beneficial to improving the accuracy of bidirectional lithium-ion battery fault detection.
  • the present invention also provides an expansion scheme for real sample data used for training, which can obtain simulated sample data based on real sample data, thereby improving the training efficiency of the fault detection model. It also eliminates the need to spend too much time on data collection during model training, which is beneficial to improving model training speed.
  • an embodiment of the present invention provides a bidirectional lithium-ion battery fault detection method based on multi-source sensing. Specifically, the above method includes the following steps:
  • Step S100 Obtain usage status data of the battery to be detected within a target time period, where the length of the target time period is a preset time period, the end time of the target time period is the current time, and the usage status data includes the first type data and second type data.
  • the first type data includes the battery heat distribution data, charge and discharge voltage data and battery temperature change data of the battery to be detected.
  • the second type data includes the battery model data and battery capacity of the battery to be detected. data, battery life data, ambient temperature data and ambient humidity data.
  • the above-mentioned battery to be tested is a bidirectional lithium-ion battery (i.e., a dual-ion battery) that requires fault detection.
  • the above-mentioned target time period is the time period for monitoring and data measurement of the battery to be tested. The length of the target time period can be based on actual needs. Settings and adjustments are not detailed here.
  • the status data collected during data collection on the battery to be tested includes first type data and second type data.
  • the first type of data is dynamic data that changes with the failure of the battery, such as battery heat distribution data, charge and discharge voltage data and battery temperature change data of the battery to be detected.
  • the second type of data is static data that will not change with the failure of the battery, such as battery model data, battery capacity data, battery usage time data, ambient temperature data, and ambient humidity data.
  • the above battery capacity data is the calibrated capacity of the battery.
  • the battery temperature change data may be the difference between the collected battery temperature and a preset temperature standard value. Or a set of continuous temperature difference data within the target time period, where a temperature difference data is the difference between the temperature value at this moment and the temperature value at the previous moment.
  • the battery temperature can be set through a Collected by temperature sensor.
  • the first type of data is obtained through multiple data collections of the battery to be tested within the target time period, and the battery thermal distribution data includes multiple heat distribution maps corresponding to the battery to be tested.
  • the above-mentioned first type of data is obtained through multiple consecutive collections of the battery to be detected, and the collection frequency can be preset and adjusted according to actual needs.
  • the value of each pixel is simulated.
  • the pixel values of the same pixel corresponding to multiple heat distribution maps have continuous values. property, sample expansion and simulation are performed based on the corresponding pixels in multiple heat distribution maps.
  • Step S200 perform principal component analysis on the above-mentioned first type data and obtain target feature data, wherein the above-mentioned target feature data is any one of the above-mentioned first type data, or the above-mentioned target feature data is composed of the above-mentioned first type data.
  • Comprehensive feature data composed of multiple data combinations.
  • the above-mentioned first type of data includes multiple different types of data, and one type of target feature data is obtained through principal component analysis processing.
  • the above-mentioned target feature data is formed by a linear combination of multiple types of data (or all types of data) in the first type of data to form a new comprehensive feature data.
  • the above-mentioned target feature data can combine the characteristics of multiple types of data.
  • dimensionality reduction can be achieved, which can not only comprehensively consider the characteristics of different data, but also reduce the amount of calculation, which is beneficial to improving the accuracy and efficiency of bidirectional lithium-ion battery fault detection.
  • Step S300 Perform wavelet transformation on the above target characteristic data to obtain target time-frequency data.
  • the above-mentioned target time-frequency data includes multiple sets of one-to-one corresponding time and frequency value data.
  • the target feature data obtained through principal component analysis processing is a one-dimensional time series signal and integrates information from multiple data. Further, perform wavelet transformation on the above target characteristic data, convert the one-dimensional time series signal into a two-dimensional time-frequency diagram, and obtain the target time-frequency data.
  • the horizontal axis of the above-mentioned target time-frequency diagram represents the time domain, and the vertical axis represents the frequency domain. .
  • the comprehensive signals corresponding to different types of fault signals are converted through wavelet transform, and the fault information characteristics are amplified, which is beneficial to further fault analysis and detection.
  • Step S400 Input the target time-frequency data and the second type data into a pre-trained fault detection model, and obtain the fault category corresponding to the battery to be detected at the current moment through the fault detection model.
  • the above-mentioned pre-trained fault detection model is a pre-trained convolutional neural network model, which is pre-trained to perform fault detection based on the input target time-frequency data and the second type of data, and output the battery to be detected at the above-mentioned current moment. Corresponding fault category.
  • the above fault categories may include one or more of temperature sensor fault, Hall reverse installation, overcharge, and BMU fault, and may also include other categories of faults, which are not specifically limited here.
  • the above fault detection model is pre-trained according to the following steps:
  • Each of the above-mentioned real sample data includes first type real data, second type real data and fault label data.
  • the above-mentioned first type The type of real data includes the battery heat distribution data, charge and discharge voltage data and battery temperature change data of the above-mentioned training battery during the training time period.
  • the above-mentioned second type of real data includes the battery model data and battery data of the above-mentioned training battery during the above-mentioned training time period. Capacity data, battery life data, ambient temperature data and ambient humidity data;
  • one of the above-mentioned simulated sample data includes first type simulation data, second type simulation data and simulation tag data.
  • the above-mentioned second type simulation data is consistent with the target real sample data.
  • the second type of real data of the sample data is the same, the above-mentioned simulated label data is the same as the fault label data of the above-mentioned target real sample data, and the above-mentioned target real sample data is the real sample data corresponding to the simulated sample data;
  • the above fault detection model is trained based on the above real sample data and the above simulated sample data.
  • multi-source information is collected from multiple training batteries of different models, and at least one of the training batteries has the same model as the battery to be processed.
  • the trained model can be applied to different types of batteries to be tested.
  • data corresponding to multiple different time periods or corresponding to different types of faults are collected.
  • training time period is the time period during which training sample data (that is, real sample data) is collected, and its time length is the same as the above target time period.
  • the sample simulation expansion can be performed to obtain multiple simulated sample data, and more sample data can be obtained for model training, thereby improving the training efficiency of the model.
  • the above-mentioned sample simulation expansion based on each of the above-mentioned real sample data to obtain multiple simulated sample data includes: for any real sample data, obtain each type of data to be processed corresponding to the above-mentioned real sample data, and according to the preset
  • the data simulation step and the statistical characteristics of the above-mentioned data to be processed perform data simulation on the above-mentioned data to be processed, obtain the target simulation data corresponding to each of the above-mentioned data to be processed, and use all the target simulation data corresponding to the above-mentioned real sample data as the real sample
  • the first type of simulation data corresponds to the simulation sample data, wherein the data to be processed includes the battery heat distribution data, charge and discharge voltage data and battery temperature change data in the real sample data.
  • the above-mentioned preset simulation step may be a preset linear combination step or a weighted average step.
  • the above-mentioned preset data simulation steps include:
  • segment statistical characteristics corresponding to each of the above-mentioned time segments wherein the above-mentioned segment statistical characteristics are the statistical characteristics of the above-mentioned data to be processed in each of the above-mentioned time segments, and the above-mentioned statistical characteristics include average value, maximum value and minimum value;
  • a set of simulated data values is generated through uniform distribution in each of the above-mentioned time segments, and we obtain Simulation processing data corresponding to the above data to be processed;
  • the above-mentioned simulation processing data and the above-mentioned data to be processed are weighted and summed according to the preset weight coefficient to obtain the target simulation data corresponding to the above-mentioned data to be processed.
  • the above-mentioned simulation processing data is data composed of all segmented simulation data values in time sequence.
  • the sum of the preset weight coefficients of the two is 1, and the specific weight ratio can be set and adjusted according to actual needs.
  • the simulated processing data and the above-mentioned data to be processed are weighted and summed. While performing data simulation, the influence of real data is fully taken into consideration, which is beneficial to improving the accuracy of the model obtained by training.
  • one of the above-mentioned real sample data corresponds to multiple simulated sample data.
  • the above-mentioned sample simulation expansion based on each of the above-mentioned real sample data to obtain multiple simulated sample data also includes: according to a preset segmentation method.
  • the data to be processed corresponding to the above-mentioned real sample data is segmented by time length and multiple different segmentation results are obtained, and a simulated sample data corresponding to the real sample data is obtained based on each segmentation result.
  • the above-mentioned preset segmentation methods include random segmentation, uniform segmentation and segmentation based on data volatility.
  • the above-mentioned training of the above-mentioned fault detection model based on the above-mentioned real sample data and the above-mentioned simulated sample data includes:
  • the above-mentioned fault detection model is trained according to the above-mentioned training data set, and the above-mentioned fault detection model is tested according to the above-mentioned test data set until the training of the above-mentioned fault detection model is completed, wherein the above-mentioned fault detection model is a convolutional god-level network model.
  • both the divided training data set and the test data set can include real sample data and simulated sample data to improve training efficiency.
  • the above-mentioned training data set includes the above-mentioned training time-frequency data and training auxiliary data.
  • the above-mentioned training auxiliary data includes second type real data, fault label data, second type simulation data and simulation label data corresponding to the above-mentioned training time-frequency data.
  • the above-mentioned test data set includes the above-mentioned test time-frequency data and test auxiliary data.
  • the above-mentioned test auxiliary data includes second type real data, fault tag data, second type simulation data and simulated tag data corresponding to the above-mentioned test time-frequency data.
  • a training time-frequency data corresponds to a training auxiliary data
  • a training auxiliary data includes a corresponding second type real data and a fault label data, representing real sample data
  • a training auxiliary data includes a corresponding second type Simulated data and a simulated tag data, representing simulated sample data.
  • a training time-frequency data corresponds to a second type of real data or a simulated label data. The same applies to the testing time-frequency data, which will not be described again here.
  • the test time-frequency data may only include to-be-processed time-frequency data corresponding to the first type of real data, that is, the test data set only corresponds to real samples to improve the accuracy of the model.
  • a corresponding set of training time-frequency data and training auxiliary data are input into the above-mentioned fault detection model.
  • the fault detection model outputs the predicted fault category. According to the predicted
  • the fault label data (or label simulation data) corresponding to the fault category and training time-frequency data adjusts the parameters of the model until the model training is completed.
  • the fault detection model training is completed.
  • a fault detection model for the bidirectional lithium-ion battery is pre-established and trained, so that the fault of the battery can be detected through the fault detection model. Specifically, perform principal component analysis on the first type of data in the usage status data corresponding to the battery to be detected and obtain the target feature data, then perform wavelet transformation to obtain the target time-frequency data, and then combine the target time-frequency data with the second type Data is used to obtain the corresponding fault category through the fault detection model.
  • there is no need to use an expert system for reasoning and a variety of different usage status data can be integrated to detect possible current faults of the bidirectional lithium-ion battery, which is beneficial to improving the accuracy of bidirectional lithium-ion battery fault detection.
  • this embodiment also provides an expansion solution for the real sample data used for training, and simulated sample data can be obtained based on the real sample data, thereby improving the training efficiency of the fault detection model. It also eliminates the need to spend too much time on data collection during model training, which is beneficial to improving model training speed.
  • Bidirectional lithium-ion battery fault detection system includes:
  • the usage status data acquisition module 510 is used to obtain the usage status data of the battery to be detected within a target time period, where the time length of the target time period is a preset time period, the end time of the target time period is the current time, and the above-mentioned usage
  • the status data includes first type data and second type data.
  • the first type data includes battery heat distribution data, charge and discharge voltage data and battery temperature change data of the battery to be detected.
  • the second type data includes battery temperature change data of the battery to be detected. Battery model data, battery capacity data, battery usage time data, ambient temperature data and ambient humidity data.
  • the first data processing module 520 is used to perform principal component analysis on the above-mentioned first type data and obtain target feature data, wherein the above-mentioned target feature data is any one of the above-mentioned first type data, or the above-mentioned target feature data is Comprehensive feature data composed of a variety of data in the above-mentioned first type of data.
  • the second data processing module 530 is used to perform wavelet transformation on the above target characteristic data to obtain target time-frequency data.
  • the fault detection module 540 is configured to input the target time-frequency data and the second type data into a pre-trained fault detection model, and obtain the fault category corresponding to the battery to be detected at the current moment through the fault detection model.
  • the above-mentioned bidirectional lithium-ion battery fault detection system based on multi-source sensing and the specific functions of each module can refer to the corresponding description in the above-mentioned bidirectional lithium-ion battery fault detection method based on multi-source sensing. Here No longer.
  • each module of the above-mentioned bidirectional lithium-ion battery fault detection system based on multi-source sensing is not unique and is not specifically limited here.
  • sequence number of each step in the above embodiment does not mean the order of execution.
  • the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
  • Module completion means dividing the internal structure of the above system into different functional units or modules to complete all or part of the functions described above.
  • Each functional unit and module in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.
  • the above-mentioned integrated unit can be hardware-based. It can also be implemented in the form of software functional units.
  • the specific names of each functional unit and module are only for the convenience of distinguishing each other and are not used to limit the scope of the present invention.
  • For the specific working processes of the units and modules in the above system please refer to the corresponding processes in the foregoing method embodiments, and will not be described again here.
  • system/intelligent terminal and method can be implemented in other ways.
  • system/intelligent terminal embodiments described above are only illustrative.
  • the division of the above modules or units is only a logical function division. In actual implementation, it can be divided in other ways, such as multiple units or units. Components may be combined or may be integrated into another system, or some features may be ignored, or not implemented.
  • the above-mentioned integrated modules/units are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned embodiment methods, and can also be completed by instructing relevant hardware through a computer program.
  • the above-mentioned computer program can be stored in a computer-readable storage medium.
  • the computer program can be stored in a computer-readable storage medium. When executed by the processor, the steps of each of the above method embodiments can be implemented.
  • the above-mentioned computer program includes computer program code, and the above-mentioned computer program code may be in the form of source code, object code, executable file or some intermediate form, etc.
  • the above-mentioned computer-readable media may include: any entity or device capable of carrying the above-mentioned computer program code, recording media, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signals, telecommunications signals, and software distribution media, etc. It should be noted that the content contained in the above computer-readable storage media can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Secondary Cells (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

La présente invention divulgue un procédé et un système de détection de défaillance de batterie à double ion basés sur une perception multi-source. Le procédé consiste : à acquérir des données d'état d'utilisation d'une batterie à tester durant une période de temps cible, la durée de la période de temps cible étant une durée prédéfinie, le moment de fin de la période de temps cible étant le moment actuel, et les données d'état d'utilisation comprenant des données d'un premier type et des données d'un second type ; à effectuer un traitement d'analyse en composantes principales sur les données du premier type et à obtenir des données caractéristiques cibles ; à effectuer une transformée en ondelettes sur les données caractéristiques cibles afin d'acquérir des données temps-fréquence cibles ; et à entrer les données temps-fréquence cibles et les données du second type dans un modèle de détection de défaillance pré-formé, et, au moyen du modèle de détection de défaillance, à acquérir une catégorie de défaillance correspondant à ladite batterie au moment actuel. Par comparaison avec l'état de la technique, la présente invention aide à améliorer la précision d'une détection de défaillance de batterie à double ion.
PCT/CN2022/136948 2022-05-27 2022-12-06 Procédé et système de détection de défaillance de batterie à double ion basés sur une perception multi-source WO2023226355A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210586138.9 2022-05-27
CN202210586138.9A CN114818831B (zh) 2022-05-27 2022-05-27 基于多源感知的双向锂离子电池故障检测方法及系统

Publications (1)

Publication Number Publication Date
WO2023226355A1 true WO2023226355A1 (fr) 2023-11-30

Family

ID=82520089

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/136948 WO2023226355A1 (fr) 2022-05-27 2022-12-06 Procédé et système de détection de défaillance de batterie à double ion basés sur une perception multi-source

Country Status (2)

Country Link
CN (1) CN114818831B (fr)
WO (1) WO2023226355A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118070023A (zh) * 2024-04-24 2024-05-24 宁德时代新能源科技股份有限公司 电池故障预测方法、训练方法、装置、设备、介质及产品
CN118656758A (zh) * 2024-08-16 2024-09-17 浙江大学 基于孪生数据多域挖掘的装备部件状态可信感知方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114818831B (zh) * 2022-05-27 2023-04-07 深圳先进技术研究院 基于多源感知的双向锂离子电池故障检测方法及系统
CN116520270B (zh) * 2023-07-04 2023-09-05 四川天中星航空科技有限公司 一种基于评估模型的雷达电子战测试方法
CN117110896B (zh) * 2023-10-24 2024-01-05 湖北工业大学 一种基于知识图谱的锂离子电池故障检测方法及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140236509A1 (en) * 2013-02-21 2014-08-21 Samsung Sdi Co., Ltd. Apparatus for detecting faults in battery system
CN111007401A (zh) * 2019-12-16 2020-04-14 国网江苏省电力有限公司电力科学研究院 一种基于人工智能的电动汽车电池故障诊断方法及设备
CN113311348A (zh) * 2021-06-09 2021-08-27 湖北工业大学 一种基于小波分解和包络谱分析的电池故障识别方法
CN113670610A (zh) * 2021-06-09 2021-11-19 广州大学 基于小波变换与神经网络的故障检测方法、系统及介质
CN113821976A (zh) * 2021-09-26 2021-12-21 中国华能集团清洁能源技术研究院有限公司 一种基于集成算法的锂电池故障诊断建模方法
CN114386537A (zh) * 2022-03-23 2022-04-22 中国华能集团清洁能源技术研究院有限公司 基于CatBoost的锂电池故障诊断方法、装置及电子设备
CN114818831A (zh) * 2022-05-27 2022-07-29 深圳先进技术研究院 基于多源感知的双向锂离子电池故障检测方法及系统

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102014222694A1 (de) * 2014-11-06 2016-05-12 Robert Bosch Gmbh Vorrichtung und Verfahren zur Erkennung eines Kontaktfehlers bei einer Batteriezelle sowie Batteriemodul, Batterie, Batteriesystem, Fahrzeug, Computerprogramm und Computerprogrammprodukt
CN108387850B (zh) * 2018-05-04 2024-05-03 金卡智能集团股份有限公司 一种基于物联网的电池监测统计系统及其方法
CN111562503B (zh) * 2020-04-07 2022-05-03 天津力神电池股份有限公司 一种锂离子电池充放电设备未做故障的分析处理方法
CN112881915A (zh) * 2021-01-18 2021-06-01 恒大新能源汽车投资控股集团有限公司 锂电池的故障识别方法、识别装置及计算机可读存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140236509A1 (en) * 2013-02-21 2014-08-21 Samsung Sdi Co., Ltd. Apparatus for detecting faults in battery system
CN111007401A (zh) * 2019-12-16 2020-04-14 国网江苏省电力有限公司电力科学研究院 一种基于人工智能的电动汽车电池故障诊断方法及设备
CN113311348A (zh) * 2021-06-09 2021-08-27 湖北工业大学 一种基于小波分解和包络谱分析的电池故障识别方法
CN113670610A (zh) * 2021-06-09 2021-11-19 广州大学 基于小波变换与神经网络的故障检测方法、系统及介质
CN113821976A (zh) * 2021-09-26 2021-12-21 中国华能集团清洁能源技术研究院有限公司 一种基于集成算法的锂电池故障诊断建模方法
CN114386537A (zh) * 2022-03-23 2022-04-22 中国华能集团清洁能源技术研究院有限公司 基于CatBoost的锂电池故障诊断方法、装置及电子设备
CN114818831A (zh) * 2022-05-27 2022-07-29 深圳先进技术研究院 基于多源感知的双向锂离子电池故障检测方法及系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118070023A (zh) * 2024-04-24 2024-05-24 宁德时代新能源科技股份有限公司 电池故障预测方法、训练方法、装置、设备、介质及产品
CN118656758A (zh) * 2024-08-16 2024-09-17 浙江大学 基于孪生数据多域挖掘的装备部件状态可信感知方法

Also Published As

Publication number Publication date
CN114818831B (zh) 2023-04-07
CN114818831A (zh) 2022-07-29

Similar Documents

Publication Publication Date Title
WO2023226355A1 (fr) Procédé et système de détection de défaillance de batterie à double ion basés sur une perception multi-source
KR102439041B1 (ko) 신경망 기반의 배터리 셀 불량 및 화재 사전 진단 방법 및 장치
CN112858919B (zh) 一种基于聚类分析的电池系统在线故障诊断方法和系统
CN111624494B (zh) 基于电化学参数的电池分析方法和系统
KR102362532B1 (ko) 신경망 기반의 배터리 잔존 수명 예측 방법 및 장치
CN111832221B (zh) 基于特征筛选的锂电池寿命预测方法
CN113554200B (zh) 动力电池电压不一致性预测方法、系统及设备
CN105911476A (zh) 一种基于数据挖掘的电池储能系统soc预测方法
CN114578234A (zh) 一种考虑因果性特征的锂离子电池退化与容量预测模型
CN116953547A (zh) 一种储能电池健康度评估方法、系统、设备及存储介质
CN115481796A (zh) 基于贝叶斯混合神经网络的电池剩余使用寿命预测方法
CN115600728A (zh) 一种动力电池的年度碳排放量估算方法及装置
CN114879046A (zh) 一种基于卡尔曼滤波的锂电池寿命预测方法和系统
CN114325433A (zh) 基于电化学阻抗谱测试的锂离子电池故障检测方法和系统
Zhang et al. Remaining useful life prediction of lithium-ion batteries based on TCN-DCN fusion model combined with IRRS filtering
CN116774064A (zh) 电池自放电检测方法、装置、设备和存储介质
Hou et al. A vehicle alarm network for high-temperature fault diagnosis of electric vehicles
CN116774053A (zh) 一种电池剩余使用寿命的预测方法、系统及存储介质
CN114722964B (zh) 基于电网频率空间和时序特征融合的数字音频篡改被动检测方法及装置
CN116449244A (zh) 基于单次放电曲线的锂电池老化轨迹早期预测方法及系统
CN113884903B (zh) 一种基于多层感知器神经网络的电池故障诊断方法
CN115219907A (zh) 一种锂电池soc估计方法、系统、介质、设备及终端
Yao et al. State of health estimation approach for Li-ion batteries based on mechanism feature empowerment
CN116068449B (zh) 一种物理模型约束深度神经网络的锂电池评估系统及方法
KR102714714B1 (ko) 신경망 기반의 배터리 내부저항 예측 방법 및 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22943542

Country of ref document: EP

Kind code of ref document: A1