A method, device and storage medium for identifying early failure of a battery

Abstract

The present invention discloses a method, device and storage medium for identifying early failure of a battery, and belongs to the field of electrochemical technology. The method includes: using AC impedance spectrum measurement technology to perform impedance testing on the battery, and recording the test data information of the battery during the test; numbering the test data with frequency, the type of test data as a feature, and each test point as a sample to construct an impedance information data set; determining the attribute data of the battery, and combining the impedance data to judge the electrochemical process in the battery charging and discharging, and establish a neural network; fitting the impedance data of the battery based on the neural network, using the neural network to divide the weights of the impedance responses of each internal interface of the battery, and judging the battery with early failure. Through the present invention, the technical problem of the difficulty of early warning of battery failure is solved.

Description

Technical Field

[0001] The present invention belongs to the field of electrochemical technology, and in particular relates to a method, a device and a storage medium for identifying early failure of a battery. Background Art

[0002] With the continuous advancement of modern science and technology, energy storage devices are facing more stringent performance and safety requirements while improving the convenience of life. The electrolyte, as the blood of the battery, is closely related to the performance of the battery. However, due to the fact that the active particles of the electrode and the electrolyte are not in phase contact, there are many interfaces between the two. Therefore, the failure of the battery is often sudden, and there are many types of failure. Therefore, how to establish an efficient, accurate and low-cost electrical failure identification method in the early stage has become a key focus of the industry.

[0003] At present, the electrochemical impedance spectroscopy (EIS) measurement technology has impedance responses to many heterogeneous interfaces inside the battery at different scanning frequencies, which truly reflects the internal state of the battery. However, the impedance spectrum of the battery is often complex, with overlapping of multiple interface processes, and the existing impedance analysis technology is often based on liquid electrolyte systems, which greatly limits the judgment of early warning of battery failure based on the AC impedance method. Summary of the invention

[0004] In order to solve the technical problem that battery failure is difficult to detect in its early stage, the present invention provides a method, device and storage medium for identifying early failure of a battery.

[0005] According to one aspect of the present invention, a method for identifying early failure of a battery is provided, comprising:

[0006] Use AC impedance spectroscopy measurement technology to perform impedance testing on the battery and record the test data information of the battery during the test;

[0007] The test data are numbered by frequency, the type of test data is characterized, and each test point is a sample to construct an impedance information data set;

[0008] Determine the battery's attribute data, and combine the impedance data to judge the electrochemical process of battery charging and discharging, and establish a neural network;

[0009] Fitting the impedance data of the battery based on a neural network, using the neural network to divide the weights of the impedance responses of various internal interfaces in the battery, and judging the batteries with early failure, wherein the various internal interfaces in the battery include interfaces inside the electrode, interfaces inside the electrolyte, and interfaces between the electrode and the electrolyte;

[0010] The battery impedance data is fitted based on a neural network, and the weights of the impedance responses of various internal interfaces of the battery are divided using the neural network, and the batteries that fail early are judged to include:

[0011] The impedance data of the battery is fitted based on the neural network, where the fitting objective function is defined as:

[0012] ;

[0013] in, w represents the frequency in the impedance data set, v i Indicates the battery voltage, Re i represents the real value of the test impedance, lm i represents the imaginary part of impedance, P i represents the phase angle, Z i represents the amplitude, t i represents the relaxation time, ab The subscripts indicate the neuron locations. q ab express a Layer Neural Network b The neuron weights;

[0014] According to the fitted battery impedance data, batteries with abnormal characteristic interface weights are selected and marked as about to fail;

[0015] The method of fitting the impedance data of the battery based on the neural network includes:

[0016] Initial value of weight , the activation function is defined as ,in, z is the variable to be input;

[0017] Define the optimization direction in each iteration process, in, X represents the weight descent gradient, w is the frequency in the impedance data set, is the weight function of the training process; is the weight function obtained by the neural network;

[0018] The weights are updated and iterated according to the calculated gradient, that is, , It is n Training No. a Tier b The neuron weights; It is n +1 training session a Tier b The neuron weights; is the learning rate;

[0019] Now bring it into the next iteration get ,in, is the set of adjusted weights in this data set; is the set of initial weights in this data set; is the learning rate of the current iteration; F (·) is the weight function obtained by the neural network;

[0020] The final error function is defined as loss=abs( ) , and the convergence condition is and To be less than 10 -3 .

[0021] Optionally, the impedance test of the battery using an AC impedance spectroscopy measurement technique includes:

[0022] During the AC impedance test, the test frequency is scanned from the high frequency area to the low frequency area, and the total number of points in the frequency range is ≥500, wherein the high frequency area represents a frequency ≥5 MHz, and the low frequency area represents a frequency ≤1 mHz.

[0023] Optionally, the test data information of the battery during the recording test process includes:

[0024] During the test, the battery voltage, test frequency, real impedance, imaginary impedance, phase angle, amplitude and relaxation time are recorded.

[0025] Optionally, the step of numbering the test data by frequency, using the type of the test data as a feature, and using each test point as a sample to construct an impedance information data set includes:

[0026] For any frequency during the test w i , there is a voltage corresponding to V i , the real part of the impedance Re i , the imaginary part of impedance lm i , Phase Angle P i ,amplitude Z i and relaxation time t i, taking the test data of each battery as a sample, collecting the battery voltage, real part of impedance, imaginary part of impedance, phase angle, amplitude and relaxation time at each frequency point, numbering them according to frequency, and w i As the column index of the dataset, the corresponding V i , Re i , lm i , P i , Z i and t i Then, the impedance information dataset is constructed as the corresponding feature under this column.

[0027] Optionally, the attribute data of the battery includes material data, geometric data, operating temperature and factory time of the battery, wherein the material data includes the type of positive and negative electrodes of the battery, quality, ratio of active materials and additives, type of electrolyte, roughness, porosity, density and thickness;

[0028] The electrochemical processes during the charge and discharge of the battery include the transfer of electrons to active materials, the exchange of electrons and ions within the active materials, and the transfer of ions from the active materials to the electrolyte and transport in the electrolyte.

[0029] Optionally, establishing a neural network includes:

[0030] One or more input neurons are used to describe the attribute data of each battery, and one or more layers of network structure are used to describe each electrochemical process, so as to establish an overall neural network, that is, the number of nodes in each layer of the neural network is m, where m is an integer multiple of the characteristic parameter, and there are n layers in total, where n is an integer multiple of the electrochemical process. The matrix representation is ( q ab ) m×n ,in, q ab Indicates a Tier b The weight represented by each neuron is calculated, and finally the classification result of the neural network is used to output whether the battery is failed or not.

[0031] According to one aspect of the present invention, there is also provided a battery early failure identification device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the battery early failure identification method as described above when executing the computer program.

[0032] According to another aspect of the present invention, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for identifying early failure of a battery as described above are implemented.

[0033] The present invention solves the problem of difficulty in early warning of battery failure. Compared with traditional battery failure warning methods, it does not require expensive high-precision sensors and has the characteristics of lower cost and higher accuracy.

[0034] The impedance data set acquisition method of the present invention is independent of the geometric shape and type of the battery, because it is applicable to a variety of batteries and has better generalization ability than traditional methods.

[0035] The multi-layer neural network used in the present invention is based on the actual electrochemical process, and the number of units and the number of layers have a good coupling interface, and the evaluation method can be corrected by supplementing more electrochemical processes in the later stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0037] Figure 1 is a flow chart of a method for identifying early failure of a battery in an embodiment of the present invention;

[0038] Figure 2 It is a schematic diagram of the basic electrochemical process of the battery;

[0039] Figure 3 This is a schematic diagram of a multi-layer neural network built based on actual electrochemical processes;

[0040] Figure 4 The results of the early failure identification method of the present invention and the screening results of the traditional method are compared. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only embodiments of a part of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0042] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0043] The terms "comprises" and "having" and any variations thereof in the specification and claims of this application are intended to cover non-exclusive inclusions. For example, a process, method, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0044] Example 1: Reference Figure 1 , Figure 1 is a flow chart of a method for identifying early failure of a battery in an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0045] S1, use AC impedance spectroscopy measurement technology to perform impedance test on the battery and record the test data information of the battery during the test;

[0046] AC impedance spectroscopy is an electrochemical measurement method that uses a small-amplitude sinusoidal potential (or current) as a disturbance signal. By applying a small-amplitude AC sinusoidal potential wave with different frequencies, the ratio of the AC potential to the current signal (the impedance of the measured system) is measured as the frequency of the sinusoidal wave changes, or the phase angle of the impedance changes as the frequency of the sinusoidal wave changes.

[0047] The embodiment of the present invention adopts AC impedance spectrum measurement technology to test the battery, applies AC signals of different frequencies to the battery, and scans from the high frequency area (≥5 MHz) to the low frequency area (≤1 mHz) during the AC impedance test. The total number of points in the frequency range is ≥500. At the same time, the response of the battery to the signal is measured, and the test data information of the battery during the test is recorded in detail, wherein the test data information includes the voltage, test frequency, real part, imaginary part, phase angle, amplitude and relaxation time of the battery at each frequency point during the test.

[0048] In the high-frequency region, the impedance spectrum mainly reflects the electron transfer process and capacitance effect inside the battery. Through the impedance data in the high-frequency region, we can preliminarily understand the electron transfer performance and capacitance characteristics inside the battery.

[0049] In the low-frequency region, the impedance spectrum mainly reflects the ion diffusion process and charge accumulation effect inside the battery. Through the impedance data in the low-frequency region, we can gain an in-depth understanding of the ion transport performance and charge accumulation inside the battery.

[0050] The total number of points in the frequency interval is ≥500, which means that at least 500 data points will be collected in the entire frequency range from the high frequency area to the low frequency area to ensure the continuity and accuracy of the test data, providing a reliable basis for subsequent data analysis and fitting.

[0051] S2, the test data are numbered by frequency, the type of test data is characterized, and each test point is a sample to construct an impedance information data set;

[0052] For any frequency during the test w i ,in, i is the index of the test frequency, w i Indicates i There is a voltage corresponding to the test frequency. V i , the real part of the impedance Re i , the imaginary part of impedance lm i , Phase Angle P i ,amplitude Z i and relaxation time t i In the AC impedance test process, the test data of a battery is a sample. The voltage, real part of impedance, imaginary part of impedance, phase angle, amplitude and relaxation time of the battery at each frequency point are collected and numbered according to the frequency. w i As the column index of the data set, the corresponding V i , Re i , lm i , P i , Z i and t i The corresponding characteristic data under this column is filled into the data set to construct an impedance information data set. In this data set, each row represents a specific frequency point. w i and its corresponding impedance data.

[0053] S3, determining the property data of the battery, and combining the impedance data to judge the electrochemical process of the battery during charging and discharging, and establishing a neural network;

[0054] The battery attribute data includes the battery's material data, geometric data, operating temperature and factory time. The material data includes the type of positive and negative electrodes, mass, ratio of active substances and additives, type of electrolyte, roughness, porosity, density and thickness. The geometric data includes the battery's size (such as length, width, height) and the area of ​​the electrodes. The operating temperature refers to the temperature of the battery when it is running. The factory time refers to the time interval from the production of the battery to the present.

[0055] During the charge and discharge process of the battery, the main electrochemical processes include the transfer of electrons to the active material (electrons are transferred from the external circuit to the active material of the battery), the exchange of electrons and ions within the active material (inside the active material, electrons and ions are exchanged to form a charge balance), and the transfer of ions from the active material to the electrolyte and transmission in the electrolyte (ions are transmitted through the electrolyte to complete the charge and discharge process of the battery).

[0056] Reference Figure 2 , Figure 2 It is a schematic diagram of the basic electrochemical process of the battery. Figure 2 As shown, Figure 2 The left side is a schematic diagram of the battery structure, from top to bottom are the positive electrode side, electrolyte and negative electrode side. Figure 2 The right side depicts the impedance data obtained from a battery test, which involves multiple charge transfer processes.

[0057] In order to establish a neural network that describes battery attribute data and electrochemical processes, the embodiment of the present invention uses one or more input neurons to describe the attribute data of each battery, and uses one or more layers of network structure to describe each electrochemical process, thereby establishing an overall neural network. The number of nodes in each layer of the established neural network is m, where m is an integer multiple of the characteristic parameter, and there are a total of n layers, where n is an integer multiple of the electrochemical process. The matrix representation form is ( q ab ) m×n , this matrix contains all the connection weights in the network, where q ab Indicates a Tier b The weights of the neurons are determined, and finally the classification result of the neural network is used to output whether the battery is failed or not.

[0058] Reference Figure 3 , Figure 3 This is a schematic diagram of a multi-layer neural network built based on an actual electrochemical process. It contains three levels: input layer, hidden layer, and output layer from left to right.

[0059] Among them, the input layer of the neural network receives the impedance data of the battery, including frequency, voltage, real part of impedance, imaginary part of impedance, phase angle, amplitude and relaxation time. These parameters constitute the input features of the neural network.

[0060] The hidden layers are responsible for learning the complex relationship between the impedance data and the internal interface impedance response. The number of hidden layers and the number of neurons in each layer can be adjusted according to the data complexity and computing resources.

[0061] The parameter output by the output layer is whether the battery is failed or not, which is equivalent to the neural network being a classifier, used to determine whether the battery is in a failed state. Traditional neural networks use data directly for calculations, but here we take the real electrochemical process into account in the neural network, turning the model from pure mathematical analysis into one with actual physical guidance.

[0062] S4, based on the neural network, fits the impedance data of the battery, uses the neural network to divide the weights of the impedance responses of each internal interface of the battery, and determines the batteries that fail early.

[0063] The established neural network is used to fit the impedance data of the battery. According to the fitting results, the weights of the impedance responses of each internal interface are obtained, wherein each internal interface includes the interface inside the electrode, the interface inside the electrolyte, and the interface between the electrode and the electrolyte. By comparing the weights of the impedance responses of each internal interface with the preset threshold, if the weight of a certain internal interface is abnormal (for example, exceeds a certain threshold), the battery is marked as about to fail, thereby classifying the batteries that fail early.

[0064] Among them, the fitting objective function is defined as:

[0065] ;

[0066] in, w represents the frequency in the impedance data set, v i Indicates the battery voltage, Re i represents the real value of the test impedance, lm i represents the imaginary part of impedance, P i represents the phase angle, Z i represents the amplitude, t i represents the relaxation time, ab The subscripts indicate the neuron locations. q ab express a Layer Neural Network b The neuron weights;

[0067] Batteries with abnormal characteristic interface weights are selected based on the fitted battery impedance data and marked as about to fail, thereby classifying batteries that fail early.

[0068] Initial value of weights during fitting , the activation function is defined as ,in, z is the variable to be input;

[0069] Define the optimization direction in each iteration process, ,in, X represents the descent gradient of the weight, w is the frequency in the impedance data set, is the weight function of the training process; is the weight function obtained by the neural network;

[0070] Then the weights are updated and iterated according to the calculated gradient, that is, , It is n Training No. a Tier b The neuron weights; It is n +1 training session a Tier b The neuron weights; is the learning rate;

[0071] Now bring it into the next iteration get ,in, is the set of adjusted weights in this data set; is the set of initial weights in this data set; is the learning rate of the current iteration; F (·) is the weight function obtained by the neural network;

[0072] The final error function is defined as loss=abs( ) , and the convergence condition is and To be less than 10 -3 .

[0073] The technical solution of the present invention is further described in detail below in conjunction with specific embodiments and drawings. It should be understood that the following embodiments are only used to explain the present invention and are not used to limit the present invention.

[0074] Example 2: In this example, the positive electrode is an NCM622 battery as the battery to be analyzed for failure, and the specific steps are as follows:

[0075] S10, connecting the battery to be analyzed to the test instrument, the open circuit voltage is detected to be 3.1 V, the AC impedance scan range is from 7 MHz to 0.001 Hz, and 500 data points are recorded;

[0076] S20, record the test data in text form, the real part of the test impedance ranges from 0.12 to 87.41Ω, the absolute value of the imaginary part ranges from 0.03 to 41.74Ω, the phase angle ranges from 11 to 27°, the amplitude is 0.04V, and the relaxation time is 0.003s;

[0077] S30, the mass of the NCM622 positive electrode was measured to be 5g, the mass fraction of active material was 70%, and the mass fraction of conductive agent was 10%; the electrolyte type was LPSC, with a porosity of 12%, a density of 2.4g / cm³, and a thickness of 0.45mm. The geometric dimensions of the battery were measured and recorded, with a length, width, and height of 50mm, 30mm, and 10mm respectively; the battery operating temperature was measured with a thermometer at 25°C, and the battery's factory life was recorded as 1700h;

[0078] S40, four semicircles were observed in the EIS test, representing four basic chemical processes: electron transfer to the cathode active material, electron and ion exchange in the active material NCM622, and ion transfer from NCM622 to the electrolyte LPSC and transport in it;

[0079] S50 uses five input neurons to describe each material information of the NCM622 battery. Each basic electrochemical process is described by a three-layer neural network to fit and weight the impedance data.

[0080] S60, after the model has been trained for one thousand times, And the rate of change of the objective function is less than 1e -3 , it is considered that the model training is completed at this time, and the training weights are saved ( q ab ) m×n ;

[0081] S70, the failure probability of the fitted NCM622 battery after 10 cycles of operation at a current density of 0.1C is 17%; the failure probability after 50 cycles of operation at a current density of 0.1C is 27%; and the failure probability after 100 cycles of operation at a current density of 0.1C is 61%.

[0082] Example 3: In this example, the positive electrode is a NCM811 battery electrode as the prediction object, and the specific steps are as follows:

[0083] S100, connect the failed battery to be analyzed to the AC impedance tester, and measure the open circuit voltage to be 3.4V. The impedance is scanned from 10MHz to 0.0001Hz, and 2000 points are taken;

[0084] S200, the test data is recorded in text form, the real part of the test impedance ranges from 0.18 to 63.51Ω, the absolute value of the imaginary part ranges from 0.021 to 31.74Ω, the phase angle ranges from 10 to 26°, the amplitude is 0.02V, and the relaxation time is 0.003s;

[0085] S300, the mass of the NCM811 positive electrode was measured to be 6g, with active material accounting for 75% and conductive carbon accounting for 8%; the electrolyte was LPSCI, with a porosity of 13%, a density of 2.6g / cm³, and a thickness of 0.4mm. The geometric dimensions of the battery were measured and recorded, with a length, width, and height of 45mm, 28mm, and 28mm, respectively. The battery operating temperature was measured with a thermometer at 23°C, and the battery's factory life was recorded as 2000h;

[0086] At S400, four semicircles were observed in the EIS test, representing four basic electrochemical processes: electron transfer to the cathode active material, electron and ion exchange in the active material NCM811, and ion transfer from NCM811 to the electrolyte LPSC and transport in it;

[0087] S500 uses 8 input neurons to describe each material information of NCM622 battery. Each basic electrochemical process is described by a four-layer neural network to fit and weight the impedance data.

[0088] S600, after the model has been trained for 800 times, And the objective function change rate is 0.4e -3 , it is considered that the model training is completed at this time, and the training weights are saved ( q ab ) m×n .

[0089] S700, the failure probability of the fitted NCM811 battery after 20 cycles of operation at a current density of 0.1C is 21%; after 50 cycles of operation at a current density of 0.1C, the failure probability is 64%.

[0090] Reference Figure 4 , Figure 4 The results of the early failure identification method of the present invention and the traditional method are compared, wherein the horizontal axis is the number of cycles to be predicted, that is, the number of charge and discharge cycles experienced by the battery, and the vertical axis is the accuracy of the prediction method, that is, the degree of consistency between the prediction result and the actual situation. Figure 4 As shown, the early failure identification method of the battery of the present invention shows higher prediction accuracy compared with the traditional method.

[0091] Embodiment 4: In this embodiment, a battery early failure identification device is also provided, which is used to implement the above embodiments and preferred implementations, and the descriptions that have been made will not be repeated. The battery early failure identification device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the battery early failure identification method described above when executing the computer program. The specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and this embodiment will not be repeated here.

[0092] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0093] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0094] 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.

[0095] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0096] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying early failure of a battery, characterized in that: include: Use AC impedance spectroscopy measurement technology to perform impedance testing on the battery and record the test data information of the battery during the test; The test data are numbered by frequency, the type of test data is characterized, and each test point is a sample to construct an impedance information data set; Determine the battery's attribute data, and combine the impedance data to judge the electrochemical process of battery charging and discharging, and establish a neural network; Fitting the impedance data of the battery based on a neural network, using the neural network to divide the weights of the impedance responses of various internal interfaces in the battery, and judging the batteries with early failure, wherein the various internal interfaces in the battery include interfaces inside the electrode, interfaces inside the electrolyte, and interfaces between the electrode and the electrolyte; The battery impedance data is fitted based on a neural network, and the weights of the impedance responses of various internal interfaces of the battery are divided using the neural network, and the batteries that fail early are judged to include: The impedance data of the battery is fitted based on the neural network, where the fitting objective function is defined as: ； in, w represents the frequency in the impedance data set, v i Indicates the battery voltage, Re i represents the real value of the test impedance, lm i represents the imaginary part of impedance, P i represents the phase angle, Z i represents the amplitude, t i represents the relaxation time, ab The subscripts indicate the neuron locations. q ab express a Layer Neural Network b The neuron weights; According to the fitted battery impedance data, batteries with abnormal characteristic interface weights are selected and marked as about to fail; The method of fitting the impedance data of the battery based on the neural network includes: Initial value of weight , the activation function is defined as ,in, z is the variable to be input; Define the optimization direction in each iteration process, ,in, X represents the weight descent gradient, w is the frequency in the impedance data set, is the weight function of the training process; is the weight function obtained by the neural network; The weights are updated and iterated according to the calculated gradient, that is, , It is n Training No. a Tier b The neuron weights; It is n +1 training session a Tier b The neuron weights; is the learning rate; Now bring it into the next iteration get ,in, is the set of adjusted weights in this data set; is the set of initial weights in this data set; is the learning rate of the current iteration; F (·) is the weight function obtained by the neural network; The final error function is defined as loss=abs( ) , and the convergence condition is and To be less than 10 -3 .

2. The method for identifying early failure of a battery according to claim 1, characterized in that: Impedance testing of batteries using AC impedance spectroscopy includes: During the AC impedance test, the test frequency is scanned from the high frequency area to the low frequency area, and the total number of points in the frequency range is ≥500, wherein the high frequency area represents a frequency ≥5 MHz, and the low frequency area represents a frequency ≤1 mHz.

3. The method for identifying early failure of a battery according to claim 1, characterized in that: The test data information of the battery during the recording test process includes: During the test, the battery voltage, test frequency, real impedance, imaginary impedance, phase angle, amplitude and relaxation time are recorded.

4. The method for identifying early failure of a battery according to claim 3, characterized in that: The test data is numbered by frequency, the type of the test data is a feature, and each test point is a sample. The impedance information data set is constructed including: For any frequency during the test w i , there is a voltage corresponding to V i , the real part of the impedance Re i , the imaginary part of impedance lm i , Phase Angle P i ,amplitude Z i and relaxation time t i , taking the test data of each battery as a sample, collecting the battery voltage, real part of impedance, imaginary part of impedance, phase angle, amplitude and relaxation time at each frequency point, numbering them according to frequency, and w i As the column index of the dataset, the corresponding V i , Re i , lm i , P i , Z i and t i Then, the impedance information dataset is constructed as the corresponding feature under this column.

5. The method for identifying early failure of a battery according to claim 1, characterized in that: The battery attribute data includes the material data, geometric data, operating temperature and factory time of the battery, wherein the material data includes the type of positive and negative electrodes, quality, ratio of active materials and additives, type of electrolyte, roughness, porosity, density and thickness of the battery; The electrochemical processes during the charge and discharge of the battery include the transfer of electrons to active materials, the exchange of electrons and ions within the active materials, and the transfer of ions from the active materials to the electrolyte and transport in the electrolyte.

6. The method for identifying early failure of a battery according to claim 1, characterized in that: The establishing of the neural network comprises: One or more input neurons are used to describe the attribute data of each battery, and one or more layers of network structure are used to describe each electrochemical process, so as to establish an overall neural network, that is, the number of nodes in each layer of the neural network is m, where m is an integer multiple of the characteristic parameter, and there are n layers in total, where n is an integer multiple of the electrochemical process. The matrix representation is ( q ab ) m×n ,in, q ab Indicates a Tier b The weight represented by each neuron is calculated, and finally the classification result of the neural network is used to output whether the battery is failed or not.

7. A battery early failure identification device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method for identifying early failure of a battery according to any one of claims 1 to 6 are implemented.

8. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for identifying early failure of a battery as claimed in any one of claims 1 to 6 are implemented.

Images