Fault early warning method, device and equipment and computer readable storage medium

Abstract

The invention relates to a fault early warning method, apparatus and device, and a computer readable storage medium. The method comprises the steps of obtaining charging data of at least one first sample and discharging data of at least one first sample; performing time domain feature extraction on the charging data of the at least one first sample to obtain at least one first feature; performing frequency domain feature extraction on the charging data of the at least one first sample to obtain at least one second feature; performing time domain feature extraction on the discharge data of the at least one first sample to obtain at least one third feature; and according to the at least one first feature, the at least one second feature and the at least one third feature, determining a target probability of fault occurrence. Therefore, the target probability of the fault can be more accurate, and the accuracy of thermal runaway early warning based on the target probability is improved.

Description

Technical Field

[0001] The present invention relates to the field of fault detection technology, and in particular to a fault early warning method, device, equipment and computer-readable storage medium. Background Art

[0002] As the world's energy is becoming increasingly scarce, green travel is becoming more and more important, and car manufacturers are starting to launch lithium battery-powered vehicles. With the increasing use of lithium battery-powered vehicles, the safety of lithium batteries has become increasingly important. Among them, the main reason for safety problems caused by thermal runaway of lithium batteries due to their own chemical reactions or external influences is that lithium batteries are prone to produce strong flames, large amounts of toxic gases, etc. after thermal runaway, causing serious personal and property losses. Therefore, when thermal runaway cannot be completely avoided, it is particularly important to improve the accuracy of thermal runaway early warning. Summary of the invention

[0003] The present invention proposes a fault warning method, device, equipment and computer-readable storage medium, which can make the target probability of a fault occurrence more accurate, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a fault warning method, which may include:

[0006] acquiring charging data of at least one first sample and discharging data of at least one first sample;

[0007] Performing time domain feature extraction on charging data of at least one first sample to obtain at least one first feature; and performing frequency domain feature extraction on charging data of at least one first sample to obtain at least one second feature;

[0008] Performing time domain feature extraction on the discharge data of at least one first sample to obtain at least one third feature;

[0009] A target probability of a failure occurring is determined based on the at least one first feature, the at least one second feature, and the at least one third feature.

[0010] According to the above technical means, based on the time domain feature extraction of the charge and discharge data, and based on the frequency domain feature extraction of the charging data, at least one first feature, at least one second feature, and at least one third feature can be obtained, so that the target probability of the fault can be obtained based on the at least one first feature, the at least one second feature, and the at least one third feature. In this way, on the one hand, the charge and discharge data can be separated for feature extraction to avoid the mutual influence between the charge and discharge data; on the other hand, the feature coverage is enhanced by extracting the frequency domain features of the charging data, so that the target probability of the fault is more accurate, thereby improving the accuracy of the thermal runaway warning based on the target probability.

[0011] In some embodiments, obtaining charging data of at least one first sample and discharging data of at least one first sample may include: obtaining at least one data, wherein the period of at least one data is greater than a first preset value; performing division processing on at least one data to obtain at least one second sample, wherein the period of the second sample is less than a second preset value, and the second preset value is less than the first preset value; determining the discharge data of at least one second sample and the charging data of at least one second sample; performing screening processing on the discharge data of at least one second sample to obtain the discharge data of at least one first sample; and performing alignment processing on the charging data of at least one second sample to obtain the charging data of at least one first sample.

[0012] According to the above technical means, on the one hand, when feature extraction is performed based on at least one second sample, since the second sample is short-period data, the number of samples can be effectively increased and the use of initial data can be reduced.

[0013] On the other hand, after screening the discharge data of at least one second sample and aligning the charging data of at least one second sample, the charging data of at least one first sample obtained is short-cycle data with the same physical interval and fixed length. When feature extraction is subsequently performed based on the discharge data of at least one first sample and the charging data of at least one first sample, the extracted features are also comparable. Based on the extracted features, the target probability of failure can also be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0014] In some embodiments, aligning the charging data of at least one second sample to obtain the charging data of at least one first sample may include: when the length of the charging data of the third sample is greater than a third preset value, intercepting the charging data of the third sample according to a preset voltage range to obtain intercepted charging data; aligning the intercepted charging data to obtain charging data of a fourth sample; wherein the third sample is one of the at least one second sample, and the fourth sample is one of the at least one first sample.

[0015] According to the above-mentioned technical means, on the one hand, when the length of the charging data of the third sample is greater than the third preset value, the charging data of the sample with too much missing charging data or too short charging time in at least one second sample is deleted, thereby improving the efficiency of aligning the charging data of at least one second sample; on the other hand, the charging data of the third sample is intercepted according to the preset voltage range, so that the voltage of the intercepted charging data can be within the preset voltage range, thereby improving the accuracy of the alignment of the intercepted charging data.

[0016] In some embodiments, aligning the intercepted charging data to obtain charging data of a fourth sample may include: when the length of the intercepted charging data is greater than a fourth preset value, eliminating the intercepted charging data to obtain charging data of the fourth sample; or, when the length of the intercepted charging data is less than or equal to the fourth preset value, interpolating the length of the intercepted charging data to obtain charging data of the fourth sample; wherein the length of the charging data of the fourth sample is equal to the fourth preset value.

[0017] According to the above technical means, based on elimination or interpolation processing, the length of the charging data of at least one first sample (for example, the fourth sample) obtained can be made equal to the fourth preset value, thereby achieving alignment of at least one first sample. When feature extraction is subsequently performed based on the charging data of at least one first sample, the extracted features are comparable, and the target probability of failure can also be accurately obtained based on the extracted features, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0018] In some embodiments, screening the discharge data of at least one second sample to obtain the discharge data of at least one first sample may include: screening the discharge data having a discharge quantity greater than a fifth preset value from the discharge data of at least one second sample to obtain the discharge data of at least one first sample.

[0019] According to the above-mentioned technical means, based on the screening of the discharge data of at least one second sample, when the discharge data of at least one first sample is subjected to feature extraction, the extracted features can be comparable, and the target probability of a fault occurring can also be accurately obtained based on the extracted features, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0020] In some embodiments, performing frequency domain feature extraction on charging data of at least one first sample to obtain at least one second feature may include: determining at least one dimension of the charging data of at least one first sample; performing frequency domain feature extraction on the charging data of at least one first sample in at least one dimension to obtain at least one second feature.

[0021] According to the above technical means, frequency domain features can be extracted from the charging data of at least one first sample in at least one dimension, and the target probability of failure can be accurately obtained based on the extracted at least one second feature, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0022] In some embodiments, frequency domain feature extraction is performed on the charging data of at least one first sample in at least one dimension to obtain at least one second feature, which may include: performing Fourier transform on the charging data of the fifth sample in the first dimension, and performing Fourier feature extraction on the Fourier transformed data to obtain at least one fourth feature; and / or performing wavelet transform on the charging data of the fifth sample in the first dimension, and performing wavelet feature extraction on the wavelet transformed data to obtain at least one fifth feature; wherein the fifth sample is one of the at least one first sample, the first dimension is one of the at least one dimension, and the at least one second feature includes at least one fourth feature and / or at least one fifth feature.

[0023] According to the above-mentioned technical means, Fourier feature extraction is performed on the charging data of the fifth sample in the first dimension, and / or wavelet feature extraction is performed on the charging data of the fifth sample in the first dimension, and based on at least one fourth feature and / or at least one fifth feature obtained, the target probability of failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0024] In some embodiments, determining a target probability of a failure based on at least one first feature, at least one second feature, and at least one third feature may include: fusing at least one first feature, at least one second feature, and at least one third feature to obtain at least one sixth feature; and inputting at least one sixth feature into the first model to obtain a target probability.

[0025] According to the above technical means, by inputting at least one sixth feature into the first model, the target probability of failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0026] In some embodiments, determining a target probability of a failure based on at least one first feature, at least one second feature, and at least one third feature may include: inputting at least one first feature, at least one second feature, and at least one third feature into a second model, a third model, and a fourth model, respectively, to obtain a first initial probability of a failure, a second initial probability of a failure, and a third initial probability of a failure, respectively; and fusing the first initial probability, the second initial probability, and the third initial probability to obtain a target probability.

[0027] According to the above technical means, the first initial probability, the second initial probability and the third initial probability are fused to accurately obtain the target probability of failure, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0028] In a second aspect, an embodiment of the present invention provides a fault warning device, the device comprising an acquisition unit, a feature extraction unit and a determination unit; wherein:

[0029] an acquisition unit configured to acquire charging data of at least one first sample and discharging data of at least one first sample;

[0030] A feature extraction unit is configured to perform time domain feature extraction on the charging data of at least one first sample to obtain at least one first feature; and perform frequency domain feature extraction on the charging data of at least one first sample to obtain at least one second feature;

[0031] The feature extraction unit is further configured to perform time domain feature extraction on the discharge data of at least one first sample to obtain at least one third feature;

[0032] The determination unit is configured to determine a target probability of a failure occurring according to at least one first feature, at least one second feature, and at least one third feature.

[0033] According to the above technical means, based on the time domain feature extraction of the charge and discharge data, and based on the frequency domain feature extraction of the charging data, at least one first feature, at least one second feature, and at least one third feature can be obtained, so that the target probability of the fault can be obtained based on the at least one first feature, the at least one second feature, and the at least one third feature. In this way, on the one hand, the charge and discharge data can be separated for feature extraction to avoid the mutual influence between the charge and discharge data; on the other hand, the feature coverage is enhanced by extracting the frequency domain features of the charging data, so that the target probability of the fault is more accurate, thereby improving the accuracy of the thermal runaway warning based on the target probability.

[0034] In a third aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising a memory and a processor, wherein:

[0035] a memory configured to store a computer program executable on the processor;

[0036] A processor is configured to execute the method as described in the first aspect when running a computer program.

[0037] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by at least one processor, it implements the method described in the first aspect.

[0038] The beneficial effects of the present invention are as follows: through the time domain feature extraction of charging data, the time domain feature extraction of discharging data and the frequency domain feature extraction of charging data, the charging and discharging data can be separated for feature extraction to avoid mutual influence between the charging and discharging data, and the feature coverage can be enhanced by extracting frequency domain features, so that when the target probability of failure is determined based on the extracted features, the accuracy of the target probability can be improved, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0039] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present invention and, together with the specification, are used to explain the technical solutions of the present invention.

[0041] Figure 1 A schematic diagram of a fault warning framework;

[0042] Figure 2 A schematic diagram of a flow chart of a fault warning method provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of a data partitioning scenario provided by an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of a scenario of discharge data and charge data in a sample provided in an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of a fault warning framework provided by an embodiment of the present invention Figure 1 ;

[0046] Figure 6 A schematic diagram of a fault warning framework provided by an embodiment of the present invention Figure 2 ;

[0047] Figure 7 A schematic diagram of a sample division and alignment process provided by an embodiment of the present invention;

[0048] Figure 8 A schematic diagram of a process for extracting frequency domain features from charging data in a single sample provided by an embodiment of the present invention;

[0049] Fig. 9 A schematic diagram of the structure of a fault warning device provided by an embodiment of the present invention;

[0050] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention;

[0051] Fig.11 A schematic diagram of the structure of another electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present invention, the implementation of the embodiments of the present invention is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not intended to limit the embodiments of the present invention.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0054] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0055] It should also be pointed out that the terms "first\second\third" involved in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present invention described here can be implemented in an order other than that illustrated or described here.

[0056] In the embodiment of the present invention, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, object A and / or object B may represent three situations: object A exists alone, object A and object B exist at the same time, and object B exists alone.

[0057] In addition, in the embodiments of the present invention, the term "at least one" may refer to one or more (two or more), unless otherwise clearly and specifically defined. For example, at least one data may refer to one or two or more data.

[0058] As the world's energy is becoming increasingly scarce, green travel is becoming more and more important, and car manufacturers are starting to launch lithium battery-powered vehicles. With the increasing use of lithium battery-powered vehicles, the safety of lithium batteries has become increasingly important. Among them, the main reason for safety problems is that lithium batteries cause thermal runaway due to their own chemical reactions or external influences. After thermal runaway, lithium batteries are prone to produce strong flames, large amounts of toxic gases, etc., causing serious personal and property losses.

[0059] At present, lithium battery-powered vehicles have good real-time data interaction capabilities. Considering the data and computing power requirements, one of the better solutions is to use cloud-based battery health management algorithms to manage the data uploaded by lithium battery-powered vehicles in real time. Cloud-based battery health management algorithms can include fault warning algorithms and life management algorithms. Among them, the lithium battery overheating warning algorithm (also known as overheating management, thermal runaway warning, etc.) is the core algorithm in the fault warning algorithm. The main technical routes of lithium battery overheating warning algorithms can include the following three categories:

[0060] The first is to use mechanism algorithms, which are based on the technical route designed based on some rule mechanism models such as damage mechanism, physical model, electrochemical model, and empirical rules to realize lithium battery overheating warning. The development of mechanism algorithms requires clarifying the cause of the failure from the mechanism end and finding the specific mechanism phenomenon, which has a high dependence on the physical and chemical systems of lithium batteries. Due to the complexity of lithium battery failure mechanisms, numerous failure modes, multiple fault coupling, and often no obvious mechanism changes before the failure occurs, the mechanism algorithms are usually complex in structure, numerous in parameters, high in mathematical complexity, and poor in versatility.

[0061] The second is to use machine learning algorithms to achieve lithium battery overheating warning through the "feature + algorithm / model" technical route. The "model" in this technical route is mainly a machine learning model, and the "feature" is one or more features extracted based on a specific fragment. This technical route is similar to the supervised learning technical route. The core of this technical route includes two parts: feature engineering (selection of data fragments and feature design) and model development (model design and training).

[0062] The third is to use an "end-to-end" approach to implement lithium battery overheating warning based on the technical route of training deep learning models based on lithium battery raw data. Although this technical route can avoid the difficulties of feature engineering, the design difficulty, training cost, data requirements, and label quality of deep learning models are still many bottlenecks that restrict this technical route.

[0063] It should be noted that from a data perspective, whether it is the development of traditional machine learning models, deep learning models or partial rule mechanism models, they are all completed on the basis of large-scale data analysis and mining. Therefore, some solutions in the above three technical routes can also be considered to be completed in a data-driven manner.

[0064] It should also be noted that the machine learning algorithm that realizes lithium battery overheating warning through the "feature + algorithm / model" technical route is currently the most important lithium battery overheating warning algorithm.

[0065] Exemplarily, the related art proposes a data-driven technical solution for lithium battery fault alarm and fault warning, which can extract corresponding value and rate data (which can be regarded as features) based on the long-cycle data of lithium batteries, where the values ​​are voltage, temperature, etc., and the rates are temperature rise rate, voltage rise rate, etc., and then similar signals / features are input into the corresponding models to realize overvoltage, undervoltage, overtemperature alarms, etc. The defect of this technical solution is that its value and rate features are too simple, and only reflect the changing laws of specific signals in the time domain, without deep mining of data features.

[0066] For example, the related technology also proposed a technical solution for predicting thermal runaway of lithium batteries based on machine learning thinking, which can realize lithium battery overheating safety warning by extracting a large number of features from the data to train a linear regression model. The defects of this technical solution are that it extracts time domain features such as voltage and temperature without deeply mining data features, charging and discharging features are not extracted separately, and the model is a single regression model. From the analysis of feature coverage and model capability, the warning accuracy of this technical solution is limited.

[0067] Figure 1 Figure 1 is a schematic diagram of a fault warning framework. Figure 1 As shown, the framework may include a data preprocessing module 110, a feature extraction module 120, a model training module 130 and a deployment test module 140, wherein:

[0068] The data preprocessing module 110 processes the long-period data and performs operations such as screening, cleaning, and deduplication on the long-period data, thereby retaining specified fields such as voltage, current, and temperature.

[0069] The feature extraction module 120 may perform time domain feature extraction on the preprocessed long-period data to obtain at least one feature.

[0070] The model training module 130 may use at least one feature as an input to a machine learning model and output the probability of a failure through the machine learning model.

[0071] By processing different long-period data through the data preprocessing module 110, the feature extraction module 120 and the model training module 130, the machine learning model can be continuously trained until a trained machine learning model is obtained.

[0072] The deployment test module 140 can obtain at least one feature after the test data passes through the data preprocessing module 110 and the feature extraction module 120, and use the at least one feature as the input of the trained machine learning model to obtain the probability of failure, so that fault warning can be performed based on the probability of failure.

[0073] Based on the above analysis, the main defects of the current machine learning algorithm are insufficient feature coverage and feature extraction only from the time domain, and the charging and discharging data are not extracted separately. When thermal runaway cannot be completely avoided, the accuracy of thermal runaway early warning is low.

[0074] Therefore, the embodiment of the present invention provides a fault warning method, which can obtain charging data of at least one first sample and discharge data of at least one first sample; extract time domain features of the charging data of at least one first sample to obtain at least one first feature; and extract frequency domain features of the charging data of at least one first sample to obtain at least one second feature; extract time domain features of the discharge data of at least one first sample to obtain at least one third feature; determine the target probability of fault occurrence according to at least one first feature, at least one second feature and at least one third feature. In this way, based on the time domain feature extraction of the charging and discharging data and the frequency domain feature extraction of the charging data, at least one first feature, at least one second feature and at least one third feature can be obtained, so that the target probability of fault occurrence can be obtained based on the at least one first feature, the at least one second feature and the at least one third feature. In this way, on the one hand, the charging and discharging data can be separated for feature extraction to avoid mutual influence between the charging and discharging data; on the other hand, the feature coverage is enhanced by extracting the frequency domain features of the charging data, so that the target probability of fault occurrence is more accurate, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0075] To facilitate understanding of the technical solution of the embodiment of the present invention, the technical solution of the present invention is described in detail below through specific embodiments. The above related technologies can be arbitrarily combined with the technical solution of the embodiment of the present invention as optional solutions, which all belong to the protection scope of the embodiment of the present invention. The embodiment of the present invention includes at least part of the following contents.

[0076] In one embodiment of the present invention, Figure 2 A schematic diagram of a fault warning method provided by an embodiment of the present invention. Figure 2 As shown, the method may include the following steps.

[0077] S210 : Acquire charging data of at least one first sample and discharging data of at least one first sample.

[0078] It should be noted that, in the embodiment of the present invention, the fault warning method can be applied to a fault warning device, or an electronic device integrated with the device. The electronic device can be implemented in various forms, for example, the electronic device can include a lithium battery powered car, a lithium battery electric car, a lithium battery mobile phone, etc., which is not limited here.

[0079] It should also be noted that the charging data may be a data segment, in which case the charging data may include at least one data; the discharging data may be a data segment, in which case the discharging data may include at least one data.

[0080] In some embodiments, obtaining charging data of at least one first sample and discharging data of at least one first sample may include: obtaining at least one data, wherein the period of at least one data is greater than a first preset value; performing division processing on at least one data to obtain at least one second sample, wherein the period of the second sample is less than a second preset value, and the second preset value is less than the first preset value; determining the discharge data of at least one second sample and the charging data of at least one second sample; performing screening processing on the discharge data of at least one second sample to obtain the discharge data of at least one first sample; and performing alignment processing on the charging data of at least one second sample to obtain the charging data of at least one first sample.

[0081] It should be noted that the at least one data may include at least one charging data and at least one discharging data.

[0082] Exemplarily, at least one data is regarded as a data segment, and the period of the data segment can be regarded as the period of the at least one data.

[0083] Exemplarily, the first preset value may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention.

[0084] It should be noted that, when the period of at least one data is greater than the first preset value, the at least one data can be considered as long-period data.

[0085] Furthermore, before acquiring at least one data, the method may further include: acquiring at least one initial data; and preprocessing the at least one initial data to obtain at least one data.

[0086] Wherein, a period of at least one initial data is greater than a first preset value.

[0087] It should be noted that preprocessing at least one initial data to obtain at least one data can be understood as screening, cleaning, deduplication and other operations on at least one initial data, so that specified fields such as voltage, current, and temperature can be retained.

[0088] Exemplarily, the division of the at least one data may be cyclic division. For example, when the at least one data is divided, the data from the end of the previous charging to the end of the current charging may be taken as a second sample.

[0089] Figure 3 A schematic diagram of a data partitioning scenario provided by an embodiment of the present invention. Figure 3 As shown, the data after the end of the n-1th charging to the end of the nth charging can be used as the nth second sample. After the n-1th charging and before the nth charging, the data needs to be discharged, left to stand, etc.

[0090] Exemplarily, the second preset value may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention.

[0091] It should be noted that, when the period of the second sample is smaller than the second preset value, and the second preset value is smaller than the first preset value, the second sample can be considered as short-period data.

[0092] Exemplarily, the second sample may be data of no more than 3 days.

[0093] It should be noted that in the related art, when feature extraction is performed based on at least one data, since at least one data is long-period data, the at least one data can extract features once. When multiple features are extracted, a large amount of data is required, which increases the computing power requirements and the efficiency of feature extraction is also low. In the embodiment of the present invention, when feature extraction is performed based on at least one second sample, since the second sample is short-period data, the number of samples can be effectively increased and the use of initial data can be reduced.

[0094] It should also be noted that in the related art, when feature extraction is performed based on at least one data, since the physical intervals of the data for feature extraction are not the same each time, the subsequently extracted features are not comparable, and the probability of a fault cannot be accurately obtained based on the extracted features; in an embodiment of the present invention, after screening the discharge data of at least one second sample and aligning the charging data of at least one second sample, the charging data of at least one first sample obtained is short-cycle data with the same physical interval and a fixed length. When feature extraction is subsequently performed based on the discharge data of at least one first sample and the charging data of at least one first sample, the extracted features are also comparable, and the target probability of a fault can also be accurately obtained based on the extracted features, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0095] In some embodiments, aligning the charging data of at least one second sample to obtain the charging data of at least one first sample may include: when the length of the charging data of the third sample is greater than a third preset value, intercepting the charging data of the third sample according to a preset voltage range to obtain intercepted charging data; aligning the intercepted charging data to obtain the charging data of the fourth sample.

[0096] The third sample is one of the at least one second sample, and the fourth sample is one of the at least one first sample.

[0097] Exemplarily, the third preset value may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention.

[0098] It should be noted that the length of the charging data of the third sample is greater than the third preset value. It can be considered that the length of the charging data of the third sample meets the minimum length requirement, so the charging data of at least one second sample with too much missing charging data or too short charging time can be deleted.

[0099] Exemplarily, the preset voltage range may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention. For example, the preset voltage range may be [V0, V1].

[0100] For example, based on specific charging rules for batteries of different models, data in which the highest module voltage of the battery in the charging segment is at [V0, V1] may be intercepted.

[0101] It should be noted that the preset voltage range needs to be large enough to cover a variety of charging states.

[0102] Through this method, on the one hand, when the length of the charging data of the third sample is greater than the third preset value, the charging data of the sample with too much missing charging data or too short charging time in at least one second sample is deleted, thereby improving the efficiency of aligning the charging data of at least one second sample; on the other hand, the charging data of the third sample is intercepted according to the preset voltage range, so that the voltage of the intercepted charging data can be within the preset voltage range, thereby improving the accuracy of the alignment of the intercepted charging data.

[0103] In some embodiments, aligning the intercepted charging data to obtain charging data of a fourth sample may include: when the length of the intercepted charging data is greater than a fourth preset value, eliminating the intercepted charging data to obtain the charging data of the fourth sample; or, when the length of the intercepted charging data is less than or equal to the fourth preset value, interpolating the length of the intercepted charging data to obtain the charging data of the fourth sample.

[0104] The length of the charging data of the fourth sample is equal to the fourth preset value.

[0105] Exemplarily, the fourth preset value may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention.

[0106] Through this method, based on elimination or interpolation processing, the length of the charging data of at least one first sample (for example, the fourth sample) obtained can be made equal to the fourth preset value, thereby achieving alignment of at least one first sample. When feature extraction is subsequently performed based on the charging data of at least one first sample, the extracted features are comparable, and the target probability of a fault can also be accurately obtained based on the extracted features, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0107] In some embodiments, screening the discharge data of at least one second sample to obtain the discharge data of at least one first sample may include: screening the discharge data having a discharge quantity greater than a fifth preset value from the discharge data of at least one second sample to obtain the discharge data of at least one first sample.

[0108] Exemplarily, the fifth preset value may be defined by a protocol or may be manually set, which is not limited in the embodiment of the present invention.

[0109] It should be noted that, when the discharge quantity of the discharge data is greater than the fifth preset value, it can be considered that the discharge data meets the specific working condition.

[0110] Through this method, based on the screening of the discharge data of at least one second sample, when the discharge data of at least one first sample is subjected to feature extraction, the extracted features can be comparable, and the target probability of a fault occurring can also be accurately obtained based on the extracted features, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0111] It should be noted that, taking the fourth sample in the at least one first sample as an example, the fourth sample may include discharge data and charge data, and the dimension of the fourth sample may include at least one of voltage, current, and temperature.

[0112] Figure 4 A schematic diagram of a scenario of discharge data and charge data in a sample provided by an embodiment of the present invention. Figure 4 As shown, assuming that the sample is the fourth sample, the length of the discharge data in the fourth sample is T1, the length of the charging data in the fourth sample is T2, the length of the data in the fourth sample is T, and the dimensions (T) of the discharge data and the charging data are voltage, current, temperature, etc.

[0113] S220, extracting time domain features from charging data of at least one first sample to obtain at least one first feature; and extracting frequency domain features from charging data of at least one first sample to obtain at least one second feature.

[0114] It should be noted that frequency domain features can reflect information that time domain features cannot reflect, such as the total energy of a signal, the proportion of low-frequency energy of a signal, and some characteristics of a signal after low-frequency / high-frequency filtering.

[0115] In some embodiments, performing frequency domain feature extraction on charging data of at least one first sample to obtain at least one second feature may include: determining at least one dimension of the charging data of at least one first sample; performing frequency domain feature extraction on the charging data of at least one first sample in at least one dimension to obtain at least one second feature.

[0116] Through this method, frequency domain features can be extracted from the charging data of at least one first sample in at least one dimension, and the target probability of failure can be accurately obtained based on the extracted at least one second feature, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0117] In some embodiments, frequency domain feature extraction is performed on the charging data of at least one first sample in at least one dimension to obtain at least one second feature, which may include: performing a Fourier transform on the charging data of the fifth sample in the first dimension, and performing Fourier feature extraction on the Fourier transformed data to obtain at least one fourth feature; and / or performing a wavelet transform on the charging data of the fifth sample in the first dimension, and performing wavelet feature extraction on the wavelet transformed data to obtain at least one fifth feature.

[0118] The fifth sample is one of the at least one first sample, the first dimension is one of the at least one dimension, and the at least one second feature includes at least one fourth feature and / or at least one fifth feature.

[0119] Exemplarily, the first dimension may be one of voltage, current, and temperature.

[0120] It should be noted that the fifth sample and the fourth sample may be the same sample or different samples, and this is not limited in the embodiment of the present invention.

[0121] It should be noted that performing Fourier transform on the charging data of the fifth sample in the first dimension may include: performing fast Fourier transform of a fixed-length time domain signal on the charging data of the fifth sample in the first dimension.

[0122] It should also be noted that performing a wavelet transform on the charging data of the fifth sample in the first dimension may include: performing a three-layer wavelet transform on the charging data of the fifth sample in the first dimension.

[0123] Exemplarily, at least one fourth feature may include at least one of the following: a DC component, a center of gravity frequency, total energy, and a proportion of energy in a certain frequency range.

[0124] Exemplarily, the at least one fifth characteristic may include energy proportion of each layer and / or capability of each layer.

[0125] Through this method, Fourier feature extraction is performed on the charging data of the fifth sample in the first dimension, and / or wavelet feature extraction is performed on the charging data of the fifth sample in the first dimension, and based on at least one fourth feature and / or at least one fifth feature obtained, the target probability of failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0126] S230: Extract time-domain features from discharge data of at least one first sample to obtain at least one third feature.

[0127] It should be noted that, based on S220 and S230, the charging data of at least one first sample can be subjected to time domain feature extraction to obtain at least one first feature; the charging data of at least one first sample can be subjected to frequency domain feature extraction to obtain at least one second feature; and the discharge data of at least one first sample can be subjected to time domain feature extraction to obtain at least one third feature. In this way, the battery charging state feature can be extracted based on the charging data alone, and the operating condition feature, driving behavior feature, etc. can be extracted based on the discharge data, so that the mutual influence between the charging and discharging data can be avoided during feature extraction.

[0128] S240. Determine a target probability of a failure occurring based on at least one first feature, at least one second feature, and at least one third feature.

[0129] In some embodiments, determining a target probability of a failure based on at least one first feature, at least one second feature, and at least one third feature may include: fusing at least one first feature, at least one second feature, and at least one third feature to obtain at least one sixth feature; and inputting at least one sixth feature into the first model to obtain a target probability.

[0130] It should be noted that the fusion of at least one first feature, at least one second feature and at least one third feature is a feature fusion.

[0131] It should also be noted that the first model can be a machine learning model.

[0132] It should also be noted that the first model may be untrained. The first model may be trained by inputting at least one sixth feature into the first model.

[0133] By using this method, at least one sixth feature is input into the first model, and the target probability of failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0134] In other embodiments, determining a target probability of a failure based on at least one first feature, at least one second feature, and at least one third feature may include: inputting at least one first feature, at least one second feature, and at least one third feature into a second model, a third model, and a fourth model, respectively, to obtain a first initial probability of a failure, a second initial probability of a failure, and a third initial probability of a failure, respectively; and fusing the first initial probability, the second initial probability, and the third initial probability to obtain a target probability.

[0135] It should be noted that the second model, the third model and the fourth model can be machine learning models.

[0136] It should also be noted that at least one of the second model, the third model, and the fourth model may be untrained. At least one of the second model, the third model, and the fourth model may be trained by inputting at least one first feature, at least one second feature, and at least one third feature into the second model, the third model, and the fourth model, respectively.

[0137] It should also be noted that the fusion of the first initial probability, the second initial probability and the third initial probability is a model fusion.

[0138] By using this method, the first initial probability, the second initial probability and the third initial probability are fused and processed, and the target probability of the failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0139] The embodiment of the present invention provides a fault warning method, which can obtain charging data of at least one first sample and discharge data of at least one first sample; extract time domain features of the charging data of at least one first sample to obtain at least one first feature; and extract frequency domain features of the charging data of at least one first sample to obtain at least one second feature; extract time domain features of the discharge data of at least one first sample to obtain at least one third feature; determine the target probability of fault occurrence according to at least one first feature, at least one second feature and at least one third feature. In this way, based on the time domain feature extraction of the charging and discharging data and the frequency domain feature extraction of the charging data, at least one first feature, at least one second feature and at least one third feature can be obtained, so that the target probability of fault occurrence can be obtained based on the at least one first feature, the at least one second feature and the at least one third feature. In this way, on the one hand, the charging and discharging data can be separated for feature extraction to avoid mutual influence between the charging and discharging data; on the other hand, the feature coverage is enhanced by extracting the frequency domain features of the charging data, so that the target probability of fault occurrence is more accurate, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0140] The fault warning method provided in the above embodiment is described in detail below in conjunction with specific application scenarios.

[0141] The embodiment of the present invention can be applied to lithium battery thermal runaway warning. The embodiment of the present invention can optimize the "feature + algorithm / model" lithium battery safety warning algorithm technical route, and proposes to extract time domain charging and discharging features and frequency domain charging features for the data samples after time domain alignment (i.e., charging data of at least one first sample and discharge data of at least one first sample), so that the extracted features can be processed by feature fusion or model fusion to achieve lithium battery overheating warning. The embodiment of the present invention can mainly include the following technical solutions:

[0142] (1) The embodiment of the present invention proposes a sample division and data alignment method. In the related art, feature extraction is performed based on cleaned long-period data, and the data range before each feature extraction is different. After sample division and data alignment, the charging feature is extracted based on short-period data samples of the same physical interval and fixed length (i.e., charging data of at least one first sample), which is both comparable and can increase the number of samples.

[0143] (2) The embodiment of the present invention proposes a frequency domain feature extraction process and method. Frequency domain features are a key supplement to the time domain feature space in the "feature + algorithm / model" technical route, which can enhance the coverage of features for fault modes and improve the accuracy of early warning.

[0144] (3) The embodiment of the present invention proposes a method for extracting charging and discharging features separately. The method of extracting various features separately and finally fusing and applying them can avoid the mutual influence between charging and discharging data during feature extraction. It can not only extract the battery charging state features based on the charging data alone, but also extract the working condition features and driving behavior features based on the discharging data. The fusion application includes feature fusion and model fusion.

[0145] The embodiment of the present invention proposes a data sample division and data alignment method, wherein the sample division and data alignment are performed after data preprocessing and before feature extraction. Figure 1 As shown in the figure, data preprocessing takes long-period data as the processing object, performs operations such as screening, cleaning, and deduplication, but does not perform sample division and alignment. There are several disadvantages in feature extraction based on this data:

[0146] (1) Features extracted from different data ranges are not comparable;

[0147] (2) Features are extracted from long-term data. A large amount of data can only extract features once. This causes the algorithm to rely on long-term data when working. A large amount of data needs to be input into the algorithm / model, which makes the model inefficient and requires high computing power.

[0148] (3) Machine learning model training requires a large amount of data. Extracting features based on long-term data also requires a large amount of data and reduces the sample size.

[0149] Based on this, in another embodiment of the present invention, Figure 5 A schematic diagram of a fault warning framework provided by an embodiment of the present invention Figure 1 .like Figure 5 As shown, the framework may include a data preprocessing module 510, a sample division and alignment module 520, a charging time domain feature extraction module 530, a discharging time domain feature extraction module 540, a charging frequency domain feature extraction module 550, a feature fusion module 560, a model training module 570 and a deployment test module 580, wherein:

[0150] The data preprocessing module 510 can process long-period data (i.e., a period greater than a first preset value) and perform operations such as screening, cleaning, and deduplication on the long-period data, thereby retaining specified fields such as voltage, current, and temperature.

[0151] The sample division and alignment module 520 can perform sample division on the preprocessed long-period data (i.e., at least one data) to obtain at least one divided sample (i.e., at least one second sample), wherein the period of the at least one divided sample is shorter (i.e., the period of the second sample is less than the second preset value), and perform data alignment on the charging data of the at least one divided sample (i.e., the charging data of the at least one second sample) to obtain the aligned charging data of the at least one sample (i.e., the charging data of the at least one first sample), and perform screening on the discharge data of the at least one divided sample (i.e., the discharge data of the at least one second sample) to obtain the screened discharge data of the at least one sample (i.e., the discharge data of the at least one first sample).

[0152] In an embodiment of the present invention, before feature extraction, the preprocessed long-period data is first divided into short-period data samples aligned with physical intervals. Each sample has the same physical interval. The features extracted on this basis are comparable to each other. At the same time, each sample is generally short-period data of no more than 3 days, thereby effectively increasing the number of samples and the amount of data input into the algorithm / model.

[0153] The charging time-domain feature extraction module 530 may perform time-domain feature extraction on the aligned charging data of at least one sample to obtain at least one first feature.

[0154] The discharge time-domain feature extraction module 540 may perform time-domain feature extraction on the discharge data of at least one sample after screening to obtain at least one third feature.

[0155] The charging frequency domain feature extraction module 550 may perform frequency domain feature extraction on the charging data of at least one aligned sample to obtain at least one second feature.

[0156] It should be noted that the FFT features and WT features can be extracted based on the Fast Fourier Transform (FFT) and 3-layer Wavelet Transform (WT) of the fixed-length time domain signal. Frequency domain features can reflect information that time domain features cannot reflect, such as the total energy of a signal, the proportion of low-frequency energy of a signal, and some features of a signal after low-frequency / high-frequency filtering.

[0157] It should also be noted that the embodiment of the present invention proposes a method of extracting charging characteristics and discharging characteristics separately, and performs different types of extraction on the charging and discharging data respectively, to ensure that the algorithm / model can pay attention to the charging status of the battery and refer to driving behavior, etc.

[0158] The feature fusion module 560 may fuse at least one first feature, at least one second feature, and at least one third feature to obtain at least one sixth feature.

[0159] The model training module 570 can use at least one sixth feature as an input to a machine learning model (ie, a first model), and output a target probability of failure through the machine learning model.

[0160] By processing different long-period data through the data preprocessing module 510, the sample division and alignment module 520, the charging time domain feature extraction module 530, the discharging time domain feature extraction module 540, the charging frequency domain feature extraction module 550 and the feature fusion module 560, the machine learning model can be continuously trained until a trained machine learning model is obtained.

[0161] The deployment test module 580 can obtain at least one feature after the test data passes through the data preprocessing module 510, the sample division and alignment module 520, the charging time domain feature extraction module 530, the discharging time domain feature extraction module 540, the charging frequency domain feature extraction module 550 and the feature fusion module 560, and use the at least one feature as the input of the trained machine learning model to obtain the probability of failure, so that fault warning can be performed based on the probability of failure.

[0162] In yet another embodiment of the present invention, Figure 6 A schematic diagram of a fault warning framework provided by an embodiment of the present invention Figure 2 .like Figure 6 As shown, the framework may include a data preprocessing module 601, a sample division and alignment module 602, a charging time domain feature extraction module 603, a discharging time domain feature extraction module 604, a charging frequency domain feature extraction module 605, a first model training module 606, a second model training module 607, a third model training module 608, a model fusion module 609 and a deployment test module 610; wherein:

[0163] The data preprocessing module 601 can take the long-period data as the processing object, and perform operations such as screening, cleaning, and deduplication on the long-period data, so as to retain the specified fields such as voltage, current, and temperature.

[0164] The sample division and alignment module 602 can perform sample division on the preprocessed long-period data to obtain at least one divided sample, wherein the period of the at least one divided sample is shorter, and perform data alignment on the charging data of the at least one divided sample to obtain the charging data of the at least one aligned sample, and perform screening on the discharge data of the at least one divided sample to obtain the discharge data of the at least one screened sample.

[0165] The charging time domain feature extraction module 603 may perform time domain feature extraction on the charging data of at least one aligned sample to obtain at least one first feature.

[0166] The discharge time-domain feature extraction module 604 may perform time-domain feature extraction on the discharge data of at least one sample after screening to obtain at least one third feature.

[0167] The charging frequency domain feature extraction module 605 may perform frequency domain feature extraction on the charging data of at least one aligned sample to obtain at least one second feature.

[0168] The first model training module 606 may use at least one first feature as an input to a first machine learning model (ie, a second model), and output a first initial probability of a failure through the first machine learning model.

[0169] The second model training module 607 can use at least one third feature as input to a third machine learning model (ie, a fourth model), and output a third initial probability of failure through the third machine learning model.

[0170] The third model training module 608 can use at least one second feature as input to a second machine learning model (ie, a third model), and output a second initial probability of failure through the second machine learning model.

[0171] By processing data of different long periods through the data preprocessing module 601, the sample division and alignment module 602, the charging time domain feature extraction module 603, the discharging time domain feature extraction module 604 and the charging frequency domain feature extraction module 605, the machine learning model (such as the first machine learning model, the second machine learning model, and the third machine learning model) can be continuously trained until a trained machine learning model is obtained.

[0172] The model fusion module 609 may fuse the first initial probability, the second initial probability and the third initial probability to obtain a target probability of a failure.

[0173] The deployment test module 610 can use the three features obtained after the test data passes through the preprocessing module 601, the sample division and alignment module 602, the charging time domain feature extraction module 603, the discharging time domain feature extraction module 604 and the charging frequency domain feature extraction module 605 as the input of the trained first machine learning model, the second machine learning model and the third machine learning model respectively. The model outputs three initial probabilities, and then the three initial probabilities are integrated to obtain the probability of failure, so that fault warning can be performed based on the probability of failure.

[0174] It should be noted that Figure 1The data preprocessing module 110, Figure 5 The data preprocessing module 510 and Figure 6 The data preprocessing module 601 in the embodiment may be the same module or different modules, which is not limited in the embodiment of the present invention.

[0175] Similarly, Figure 1 , Figure 5 and Figure 6 Some modules in other modules in the embodiment may be the same module or different modules, and the embodiments of the present invention will not be described in detail here.

[0176] It should also be noted that Figure 5 The three types of features (i.e., at least one first feature, at least one second feature, and at least one third feature) are fused to form a fused feature (i.e., at least one sixth feature) to train a single machine learning model; Figure 6 It is to use three types of features to train a machine learning model respectively, that is, to use at least one first feature to train a first machine learning model, to use at least one second feature to train a second machine learning model, and to use at least one third feature to train a third machine learning model, and then fuse the three machine learning models. It has been verified that both methods can achieve lithium battery overheating safety warning, and compared with the solution without sample division and alignment, separate extraction of charge and discharge features, and frequency domain feature extraction, both methods have obvious indicator improvements.

[0177] In another embodiment of the present invention, Figure 7 A schematic diagram of a sample division and alignment process provided by an embodiment of the present invention. Figure 7 As shown, the method may include the following steps.

[0178] S701, preprocessed data.

[0179] Data preprocessing takes long-period data as the processing object, and performs operations such as screening, cleaning, and deduplication on the long-period data, so that specified fields such as voltage, current, and temperature can be retained.

[0180] S702: Circular division.

[0181] S703, n samples to be processed.

[0182] By performing cyclic division on the preprocessed data, n samples to be processed can be obtained, and the periods of the n samples to be processed are all short periods.

[0183] Exemplarily, the range of samples to be processed is as follows Figure 3 As shown, the data from after the end of the n-1th charging to the end of the nth charging is taken as a data sample to be processed.

[0184] S704: Take a sample to be processed (ie, the third sample).

[0185] S705: Extract discharge data.

[0186] Extract discharge data from a sample to be processed.

[0187] S706: Whether the amount of power discharged is greater than a fifth preset value.

[0188] It should be noted that, when the discharge quantity of the discharge data is greater than the fifth preset value, it can be considered that the discharge data meets the specific working condition.

[0189] It should also be noted that when the discharge capacity is greater than the fifth preset value, the discharge data with a discharge capacity greater than the fifth preset value can be screened out from the discharge data, that is, S707 is executed; when the discharge capacity is less than or equal to the fifth preset value, a new sample to be processed can be taken, that is, S704 is executed.

[0190] S707: Data screening.

[0191] The discharge data having a discharge quantity greater than a fifth preset value is screened out from the discharge data.

[0192] S708: extract charging data.

[0193] Extract charging data from a sample to be processed.

[0194] S709: Whether the length of the charging data is greater than a third preset value.

[0195] It should be noted that judging whether the length of the charging data is greater than the third preset value can be understood as judging whether the length of the charging data meets the minimum length requirement. Data whose length of the charging data is less than or equal to the third preset value can be deleted, so that data with too much missing charging data or too short charging time can be deleted.

[0196] It should also be noted that when the length of the charging data is greater than the third preset value, based on the specific charging rules of different types of lithium batteries, the charging data in which the highest battery module voltage in the charging segment is in the interval [V0, V1] (i.e., the preset voltage range) can be intercepted to obtain the intercepted charging data, and then it is determined again whether the length of the intercepted charging data is greater than the fourth preset value, so as to determine whether to perform uniform elimination or linear interpolation processing on the intercepted charging data, that is, execute S710; when the length of the charging data is less than or equal to the third preset value, a new sample to be processed can be taken, that is, execute S704.

[0197] Exemplarily, when the voltage interval [V0, V1] is large enough, multiple charging states can be covered.

[0198] S710: Whether the length of the charging data is greater than a fourth preset value.

[0199] It should be noted that the charging data at this time is intercepted charging data.

[0200] It should also be noted that when the length of the charging data is greater than the fourth preset value, the charging data can be uniformly eliminated, that is, S711 is executed; when the length of the charging data is less than or equal to the fourth preset value, the charging data can be linearly interpolated, that is, S712 is executed.

[0201] S711, uniform elimination.

[0202] The charging data is evenly eliminated to ensure that the length of the charging data (ie, the charging data of the fourth sample) is equal to the fourth preset value.

[0203] S712. Linear interpolation.

[0204] Linear interpolation is performed on the charging data to determine that the length of the charging data is equal to a fourth preset value.

[0205] Through this method, on the one hand, long-period data are divided into short-period samples to be processed. When feature extraction is performed based on the short-period samples to be processed, the number of samples can be effectively increased and the use of initial data can be reduced; on the other hand, the screening of discharge data based on the samples to be processed can make the extracted features comparable when feature extraction is performed on the screened discharge data; on the other hand, the elimination or interpolation processing of the charging data of the samples to be processed can make the length of the processed charging data equal to the fourth preset value, thereby realizing the alignment of the charging data. When feature extraction is subsequently performed based on the charging data, the extracted features can be comparable.

[0206] It should be noted that after the sample division and alignment, n T×D samples can be obtained, each sample is as follows Figure 4 shown.

[0207] In yet another embodiment of the present invention, Figure 8 A schematic diagram of a process for extracting frequency domain features from charging data in a single sample provided by an embodiment of the present invention. Figure 8As shown, the charging data in a single sample is a subset of all the data in a single sample, which is a T2×D data segment, and each dimension is a single-dimensional time series data of T2×1; 1 dimension can extract 1 set of frequency domain features, so D dimensions can extract at most D sets of frequency domain features. 1 set of frequency domain features can include m1 features extracted based on FFT transform and m2 features extracted based on WT transform, and at most (m1+m2)*D frequency domain features can be extracted. In practical applications, (m1+m2)*D features can be screened, and only the m features with the highest degree of influence on the model are retained. The method may include the following steps.

[0208] S801 , obtaining single-dimensional (ie, first-dimensional) charging data in a single sample (ie, the fifth sample).

[0209] It should be noted that obtaining single-dimensional charging data in a single sample can be understood as obtaining a single-dimensional time-series charging fragment (data in the time domain).

[0210] Exemplarily, the single dimension may be voltage.

[0211] It should also be noted that the single sample may be a sample that has been divided and aligned.

[0212] S802: Determine whether it is necessary to perform time domain processing on the single-dimensional charging data.

[0213] Determine whether the single-dimensional charging data needs to be processed in the time domain, such as if some signals need to be converted by integration, range finding, etc. If the single-dimensional charging data needs to be processed in the time domain, the single-dimensional charging data can be processed in the time domain (such as integral conversion, and range finding conversion) to determine the new single-dimensional charging data in a single book, that is, execute S803; if the single-dimensional charging data does not need to be processed in the time domain, the single-dimensional charging data can be FFT transformed, and / or the single-dimensional charging data can be WT transformed, that is, execute S804 and / or S807.

[0214] S803: Determine new single-dimensional charging data in a single sample.

[0215] The one-dimensional charging data is transformed by integration, range finding and other operations to determine new one-dimensional charging data in a single sample.

[0216] S804: Perform FFT transformation on the single-dimensional charging data.

[0217] S805 , performing FFT feature extraction on the single-dimensional charging data after FFT transformation.

[0218] S806: Determine FFT-type features (ie, at least one fourth feature).

[0219] It should be noted that when performing FFT feature extraction on the single-dimensional charging data after FFT transformation, the FFT-type features can be calculated using a specific formula.

[0220] Exemplarily, the FFT-type features may include at least one of the following: a direct current component, a center of gravity frequency, a total energy, and an energy proportion of a certain frequency band.

[0221] S807 , performing WT transformation on the single-dimensional charging data.

[0222] S808 , extract WT features from the single-dimensional charging data after WT transformation.

[0223] S809: Determine WT-type features (ie, at least one fifth feature).

[0224] It should be noted that when WT feature extraction is performed on the single-dimensional charging data after WT transformation, the WT-type features can be calculated through a specific formula.

[0225] Exemplarily, the WT-type features may include the energy proportion of each layer and / or the capability of each layer.

[0226] Through this method, FFT feature extraction is performed on the single-dimensional charging data after FFT transformation, and / or WT feature extraction is performed on the single-dimensional charging data after WT transformation. Based on the obtained FFT-type features and / or WT-type features, the target probability of failure can be accurately obtained, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0227] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the present invention. For example, the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. For another example, the various different embodiments of the present invention can also be arbitrarily combined, as long as they do not violate the idea of ​​the present invention, they should also be regarded as the contents disclosed by the present invention. For another example, under the premise of no conflict, the various embodiments described in the present invention and / or the technical features in each embodiment can be arbitrarily combined with the prior art, and the technical solution obtained after the combination should also fall within the protection scope of the present invention.

[0228] It should also be understood that in the various method embodiments of the present invention, the size of the serial numbers of the above-mentioned processes 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 embodiments of the present invention.

[0229] In yet another embodiment of the present invention, based on the same inventive concept as the above embodiment, see Fig. 9 , which shows a schematic diagram of the structure of the fault warning device provided by an embodiment of the present invention. Fig. 9 As shown, the fault warning device 900 may include an acquisition unit 910, a feature extraction unit 920 and a determination unit 930; wherein:

[0230] An acquisition unit 910, configured to acquire charging data of at least one first sample and discharging data of at least one first sample;

[0231] The feature extraction unit 920 is configured to perform time domain feature extraction on the charging data of at least one first sample to obtain at least one first feature; and perform frequency domain feature extraction on the charging data of at least one first sample to obtain at least one second feature; perform time domain feature extraction on the discharge data of at least one first sample to obtain at least one third feature;

[0232] The determination unit 930 is configured to determine a target probability of a failure occurring according to at least one first feature, at least one second feature, and at least one third feature.

[0233] In some embodiments, the fault warning device 900 may further include a processing unit 940, wherein:

[0234] The acquisition unit 910 is further configured to acquire at least one data, wherein the period of the at least one data is greater than a first preset value;

[0235] The processing unit 940 is configured to perform division processing on at least one data to obtain at least one second sample, wherein the period of the second sample is less than a second preset value, and the second preset value is less than the first preset value;

[0236] The determination unit 930 is further configured to determine discharge data of at least one second sample and charge data of at least one second sample;

[0237] The processing unit 940 is further configured to perform screening processing on the discharge data of at least one second sample to obtain the discharge data of at least one first sample; and perform alignment processing on the charge data of at least one second sample to obtain the charge data of at least one first sample.

[0238] In some embodiments, the processing unit 940 is further configured to, when the length of the charging data of the third sample is greater than a third preset value, intercept the charging data of the third sample according to a preset voltage range to obtain intercepted charging data; align the intercepted charging data to obtain charging data of a fourth sample; wherein the third sample is one of the at least one second sample, and the fourth sample is one of the at least one first sample.

[0239] In some embodiments, the processing unit 940 is further configured to, when the length of the intercepted charging data is greater than a fourth preset value, eliminate the intercepted charging data to obtain charging data of a fourth sample; or, when the length of the intercepted charging data is less than or equal to the fourth preset value, interpolate the length of the intercepted charging data to obtain charging data of a fourth sample; wherein the length of the charging data of the fourth sample is equal to the fourth preset value.

[0240] In some embodiments, the fault warning device 900 may further include a screening unit 950, wherein:

[0241] The screening unit 950 is configured to screen out discharge data with a discharge quantity greater than a fifth preset value from the discharge data of at least one second sample, and obtain the discharge data of at least one first sample.

[0242] In some embodiments, the determination unit 930 is further configured to determine at least one dimension of the charging data of at least one first sample; the feature extraction unit 920 is further configured to perform frequency domain feature extraction on the charging data of at least one first sample in at least one dimension to obtain at least one second feature.

[0243] In some embodiments, the feature extraction unit 920 is further configured to perform a Fourier transform on the charging data of the fifth sample in the first dimension, and perform Fourier feature extraction on the Fourier transformed data to obtain at least one fourth feature; and / or perform a wavelet transform on the charging data of the fifth sample in the first dimension, and perform wavelet feature extraction on the wavelet transformed data to obtain at least one fifth feature; wherein the fifth sample is one of the at least one first sample, the first dimension is one of the at least one dimension, and the at least one second feature includes at least one fourth feature and / or at least one fifth feature.

[0244] In some embodiments, the processing unit 940 is further configured to perform a fusion process on at least one first feature, at least one second feature, and at least one third feature to obtain at least one sixth feature; and input the at least one sixth feature into the first model to obtain a target probability.

[0245] In some embodiments, the processing unit 940 is further configured to input at least one first feature, at least one second feature, and at least one third feature into the second model, the third model, and the fourth model, respectively, to obtain a first initial probability of failure, a second initial probability of failure, and a third initial probability of failure, respectively; and to fuse the first initial probability, the second initial probability, and the third initial probability to obtain a target probability.

[0246] It can be understood that in this embodiment, a "unit" can be a part of a processor, a part of a program or software, etc., and of course, it can also be a module, or it can be non-modular. Moreover, the components in this embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional module.

[0247] If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0248] Therefore, this embodiment provides a computer storage medium, which stores a computer program. When the computer program is executed by at least one processor, the steps of the method described in any one of the above embodiments are implemented.

[0249] Based on the above-mentioned components of the fault warning device 900 and the computer storage medium, see Fig.10 , which shows a schematic diagram of the composition structure of an electronic device provided by an embodiment of the present invention. Fig.10As shown, the electronic device 1000 may include: a communication interface 1010, a memory 1020 and a processor 1030; each component is coupled together via a bus system 1040. It is understood that the bus system 1040 is configured to achieve connection and communication between these components. In addition to the data bus, the bus system 1040 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Fig.10 Various buses are labeled as bus system 1040. Among them, the communication interface 1010 is configured to receive and send signals during the process of sending and receiving information between other external devices or applications;

[0250] A memory 1020 configured to store a computer program that can be executed on the processor 1030;

[0251] The processor 1030 is configured to execute, when running the computer program:

[0252] acquiring charging data of at least one first sample and discharging data of at least one first sample;

[0253] Performing time domain feature extraction on charging data of at least one first sample to obtain at least one first feature; and performing frequency domain feature extraction on charging data of at least one first sample to obtain at least one second feature;

[0254] Performing time domain feature extraction on the discharge data of at least one first sample to obtain at least one third feature;

[0255] A target probability of a failure occurring is determined based on the at least one first feature, the at least one second feature, and the at least one third feature.

[0256] It can be understood that the memory 1020 in the embodiment of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRRAM). The memory 1020 of the system and method described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0257] The processor 1030 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 1030. The above processor 1030 may be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined and executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 1020, and the processor 1030 reads the information in the memory 1020 and completes the steps of the above method in combination with its hardware.

[0258] It is understood that the embodiments described in the present invention can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present invention or a combination thereof.

[0259] For software implementation, the technology described in the present invention can be implemented by a module (such as a procedure, function, etc.) that performs the functions described in the present invention. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0260] Optionally, as another embodiment, the processor 1030 is further configured to execute the steps of the method described in any one of the aforementioned embodiments when running the computer program.

[0261] Based on the above-mentioned components of the fault warning device 900 and the computer storage medium, see Fig.11 , which shows a schematic diagram of the composition structure of another electronic device provided by an embodiment of the present invention. Fig.11 As shown, the electronic device 1000 may include the fault warning device 900 described in any one of the aforementioned embodiments.

[0262] In an embodiment of the present invention, charging data of at least one first sample and discharging data of at least one first sample can be obtained; time domain feature extraction is performed on the charging data of at least one first sample to obtain at least one first feature; and frequency domain feature extraction is performed on the charging data of at least one first sample to obtain at least one second feature; time domain feature extraction is performed on the discharging data of at least one first sample to obtain at least one third feature; and the target probability of failure is determined according to at least one first feature, at least one second feature and at least one third feature. In this way, based on the time domain feature extraction of the charging and discharging data and the frequency domain feature extraction of the charging data, at least one first feature, at least one second feature and at least one third feature can be obtained, so that the target probability of failure can be obtained based on the at least one first feature, the at least one second feature and the at least one third feature. In this way, on the one hand, the charging and discharging data can be separated for feature extraction to avoid mutual influence between the charging and discharging data; on the other hand, the feature coverage is enhanced by extracting the frequency domain features of the charging data, so that the target probability of failure is more accurate, thereby improving the accuracy of thermal runaway warning based on the target probability.

[0263] It should be noted that, in the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0264] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0265] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0266] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0267] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0268] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A fault warning method, characterized in that: The method comprises: acquiring charging data of at least one first sample and discharging data of the at least one first sample; Performing time domain feature extraction on the charging data of the at least one first sample to obtain at least one first feature; and performing frequency domain feature extraction on the charging data of the at least one first sample to obtain at least one second feature; Performing time domain feature extraction on the discharge data of the at least one first sample to obtain at least one third feature; A target probability of a failure occurring is determined based on the at least one first feature, the at least one second feature, and the at least one third feature.

2. The method according to claim 1, characterized in that The acquiring charging data of at least one first sample and discharging data of at least one first sample comprises: Acquire at least one data, wherein a period of the at least one data is greater than a first preset value; Performing division processing on the at least one data to obtain at least one second sample, wherein a period of the second sample is smaller than a second preset value, and the second preset value is smaller than the first preset value; determining discharge data of the at least one second sample and charge data of the at least one second sample; Performing screening processing on the discharge data of the at least one second sample to obtain the discharge data of the at least one first sample; Alignment processing is performed on the charging data of the at least one second sample to obtain the charging data of the at least one first sample.

3. The method according to claim 2, characterized in that The aligning the charging data of the at least one second sample to obtain the charging data of the at least one first sample includes: When the length of the charging data of the third sample is greater than the third preset value, intercepting the charging data of the third sample according to the preset voltage range to obtain intercepted charging data; Performing alignment processing on the intercepted charging data to obtain charging data of a fourth sample; The third sample is one of the at least one second sample, and the fourth sample is one of the at least one first sample.

4. The method according to claim 3, characterized in that The aligning the intercepted charging data to obtain charging data of a fourth sample includes: When the length of the intercepted charging data is greater than a fourth preset value, the intercepted charging data is eliminated to obtain the charging data of the fourth sample; or When the length of the intercepted charging data is less than or equal to the fourth preset value, interpolating the length of the intercepted charging data to obtain the charging data of the fourth sample; The length of the charging data of the fourth sample is equal to the fourth preset value.

5. The method according to any one of claims 2 to 4, characterized in that: The screening process of the discharge data of the at least one second sample to obtain the discharge data of the at least one first sample includes: Discharge data with a discharge quantity greater than a fifth preset value is screened out from the discharge data of the at least one second sample to obtain the discharge data of the at least one first sample.

6. The method according to any one of claims 1 to 4, characterized in that The extracting frequency domain features of the charging data of the at least one first sample to obtain at least one second feature includes: determining at least one dimension of charging data of the at least one first sample; Frequency domain feature extraction is performed on the charging data of the at least one first sample in the at least one dimension to obtain the at least one second feature.

7. The method according to claim 6, characterized in that The performing frequency domain feature extraction on the charging data of the at least one first sample in the at least one dimension to obtain the at least one second feature includes: Performing Fourier transformation on the charging data of the fifth sample in the first dimension, and performing Fourier feature extraction on the Fourier transformed data to obtain at least one fourth feature; and / or, performing a wavelet transform on the charging data of the fifth sample in the first dimension, and performing a wavelet feature extraction on the wavelet transformed data to obtain at least one fifth feature; The fifth sample is one of the at least one first sample, the first dimension is one of the at least one dimension, and the at least one second feature includes the at least one fourth feature and / or the at least one fifth feature.

8. The method according to any one of claims 1 to 4, characterized in that The determining, according to the at least one first feature, the at least one second feature, and the at least one third feature, a target probability of a failure occurring comprises: fusing the at least one first feature, the at least one second feature, and the at least one third feature to obtain the at least one sixth feature; The at least one sixth feature is input into the first model to obtain the target probability.

9. The method according to any one of claims 1 to 4, characterized in that The determining, according to the at least one first feature, the at least one second feature, and the at least one third feature, a target probability of a failure occurring comprises: Inputting the at least one first feature, the at least one second feature and the at least one third feature into the second model, the third model and the fourth model respectively, to obtain the first initial probability of failure, the second initial probability of failure and the third initial probability of failure respectively; The first initial probability, the second initial probability and the third initial probability are fused to obtain the target probability.

10. A fault warning device, characterized in that: The device comprises an acquisition unit, a feature extraction unit and a determination unit; wherein: The acquisition unit is configured to acquire charging data of at least one first sample and discharging data of the at least one first sample; The feature extraction unit is configured to perform time domain feature extraction on the charging data of the at least one first sample to obtain at least one first feature; and perform frequency domain feature extraction on the charging data of the at least one first sample to obtain at least one second feature; The feature extraction unit is further configured to perform time domain feature extraction on the discharge data of the at least one first sample to obtain at least one third feature; The determination unit is configured to determine a target probability of a failure occurring based on the at least one first feature, the at least one second feature, and the at least one third feature.

11. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein: The memory is configured to store a computer program that can be executed on the processor; The processor is configured to execute the method according to any one of claims 1 to 9 when running the computer program.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by at least one processor, the method according to any one of claims 1 to 9 is implemented.

Images