阅读说明：本技术 一种锂离子电池故障诊断方法 (Lithium ion battery fault diagnosis method ) 是由 曲杰 甘伟 于 2020-04-23 设计创作，主要内容包括：本发明公开了一种锂离子电池故障诊断方法,包括以下步骤：<I>S</I>1、获取锂离子电池未发生故障及发生各类型故障时的电池数据；<I>S</I>2、使用信号降噪模型对步骤<I>S</I>1得到的数据进行降噪处理；<I>S</I>3、通过特征提取模型计算得到表征锂离子电池内部不同频率化学反应的特征参数；<I>S</I>4、计算该锂离子电池的安全阈值；<I>S</I>5、根据所述安全阈值确定报警时对应的故障类型,建立锂离子电池故障诊断模型；<I>S</I>6、获取待诊断锂离子电池使用过程中充放电循环电池数据,对该数据降噪并计算特征参数曲线,将特征参数曲线与基准参数曲线进行相似度对比,得到相似程度；<I>S</I>7、将相似程度输入步骤<I>S</I>5的锂离子电池故障诊断模型,若达到阈值,发出相应故障类型报警信号。(The invention discloses a lithium ion battery fault diagnosis method, which comprises the following steps: S 1. acquiring battery data when the lithium ion battery fails and various types of faults occur; S 2. step pair using signal noise reduction model S 1, carrying out noise reduction processing on the data obtained; S 3. calculating to obtain characteristic parameters representing chemical reactions with different frequencies in the lithium ion battery through a characteristic extraction model; S 4. calculating the safety threshold of the lithium ion battery; S 5. determining a corresponding fault type during alarming according to the safety threshold, and establishing a lithium ion battery fault diagnosis model; S 6. acquiring charge-discharge cycle battery data in the use process of the lithium ion battery to be diagnosed, denoising the data, calculating a characteristic parameter curve, and comparing the similarity of the characteristic parameter curve and a reference parameter curve to obtain a similarity rangeDegree; S 7. inputting the degree of similarity into the step S And 5, if the lithium ion battery fault diagnosis model reaches a threshold value, sending a corresponding fault type alarm signal.)

1. A lithium ion battery fault diagnosis method is characterized by comprising the following steps:

s1, acquiring battery data when the lithium ion battery fails and various types of faults occur;

s2, carrying out noise reduction processing on the data obtained in the step S1 by using a signal noise reduction model;

s3, calculating characteristic parameters representing chemical reactions of different frequencies in the lithium ion battery through a characteristic extraction model;

s4, calculating the safety threshold of the lithium ion battery by comparing the similarity of the characteristic parameter curves of the lithium ion battery without faults and the lithium ion battery with various types of faults;

s5, determining the corresponding fault type during alarming according to the safety threshold, and establishing a lithium ion battery fault diagnosis model;

s6, acquiring charge and discharge cycle battery data in the use process of the lithium ion battery to be diagnosed, denoising the data and calculating to obtain a characteristic parameter curve, and comparing the similarity of the characteristic parameter curve and a reference parameter curve to obtain the similarity;

and S7, inputting the similarity degree into the lithium ion battery fault diagnosis model in the step S5, and if the similarity degree reaches a threshold value, sending out a corresponding fault type alarm signal.

2. The method according to claim 1, wherein the battery data obtained in step S1 includes battery voltage, probe temperature and corresponding fault type in the lithium ion battery history data.

3. The method for diagnosing faults of a lithium ion battery according to claim 1, wherein the noise reduction processing of the signal noise reduction model in step S2 includes the following steps:

converting the obtained lithium ion battery data from the time series s (t) into a time-frequency domain, i.e. Continuous Wavelet Transform (CWT) coefficients, using a Continuous Wavelet Transform (CWT), the obtained time-frequency domain being:

where α denotes the continuous wavelet analysis scale, τ denotes the continuous wavelet analysis time transfer window size, t denotes time, denotes the complex conjugate,<s,ψ_α,τ>denotes s (t) andthe inner product of (a) is,representing an analysis function in a continuous wavelet transform, i.e. a parent wave;

the resulting time-frequency domain is divided into a high-energy low-frequency part and a high-energy high-frequency part:

wherein, W_s(α, τ) is the resulting time-frequency domain, n_aFor the number of scales, CF (tau) is the superposition amplitude of the CWT coefficient calculated by using all continuous wavelet analysis scales a, the distribution of the CWT coefficient along the scale axis is obtained by the CF coefficient, because the existence of the low-frequency characteristic enables the obtained distribution to have two different peak values, and the two different peak values are divided by setting an optimal threshold value, so that a high-energy low-frequency part and a high-energy high-frequency part are obtained;

synchronously compressing the low-frequency part and the high-frequency part by using synchronous compression continuous wavelet transform (SS-CWT) respectively to obtain corresponding instant frequency, namely a synchronous compression continuous wavelet transform coefficient (SS-CWT) coefficient, wherein the instant frequency is as follows:

i represents a complex number, representing a partial derivative;

for the instantaneous frequency obtained by synchronous compression of the low-frequency part and the instantaneous frequency obtained by synchronous compression of the high-frequency part, different methods are adopted for noise reduction, and the method specifically comprises the following steps:

for a low-frequency part, introducing a soft interval screening characteristic to filter noise, wherein the soft interval is as follows:

lambda is the set threshold value and is the threshold value,indicates the characteristics after screening, omega_sRepresenting continuous wavelet transform coefficients;

for the high frequency part, calculating the superposed amplitude CF of the front signal segment, screening the characteristics of the high frequency part by using a hard interval, and filtering the main noise, wherein the hard interval is as follows:

wherein λ is_nTo set the threshold value, M_max＝mean(max|T_nI), and T)_rRespectively calculating SS-CWT coefficients of narrow frequency bands corresponding to the two peak values obtained after CF calculation;

combining the SS-CWT coefficient of the high-frequency part and the SS-CWT coefficient of the low-frequency part after noise reduction to form a time-frequency domain after noise reduction, and inverting to a time sequence signal;

and (3) converting the denoised time sequence signal again through continuous wavelet transform to obtain a Continuous Wavelet Transform (CWT) coefficient, and performing post-denoising again by using a CT threshold, wherein the CT threshold is as follows:

where λ is the set threshold, 0<γ<Lambda is more than or equal to 0 and less than or equal to α and less than or equal to 1, gamma is a truncation value, when the continuous wavelet transform coefficient is less than the truncation value, the coefficient is set to 0,for the selected characteristics, sgn (W)_s) Indicating that if the continuous wavelet transform coefficient is positive, the output is 1, otherwise-1, W is output_sRepresenting continuous wavelet transform coefficients;

and the signal denoising model outputs a Continuous Wavelet Transform (CWT) coefficient obtained by the post-denoising.

4. The lithium ion battery fault diagnosis method according to claim 1, wherein step S3 specifically includes:

calculating a parameter DF by multi-scale envelope superposition of Continuous Wavelet Transform (CWT) coefficients output by a signal denoising module, wherein DF is:

where n is the number of scales, E (α, tau) is the envelope function of the Continuous Wavelet Transform (CWT) coefficients, and the formula isHere, theIs the result of the hilbert transform of the continuous wavelet transform coefficients;

calculating an operating energy ratio ER using the DF obtained above₁Said operating energy ratio ER₁Comprises the following steps:

wherein L is the length of the energy collection window before and after the time transfer window size τ;

using ER obtained as described above₁Calculating characteristic parameter ER for representing chemical reactions with different frequencies in lithium ion battery₂The ER₂Comprises the following steps:

ER₂(τ)＝ER₁(τ)|DF(α)|

the characteristic extraction model outputs the characteristic parameter ER obtained above₂。

5. The method for diagnosing the lithium ion battery fault according to claim 1, wherein the step S4 of calculating the safety threshold of the lithium ion battery specifically includes:

the short-circuit threshold calculation unit is used for inputting the charging cycle voltage data of the lithium ion battery without fault and the charging cycle voltage data of the lithium ion battery with short-circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the charging cycle voltage data and the charging cycle voltage data to obtain a safety threshold of the short-circuit fault;

the open circuit threshold calculation unit is used for inputting the discharge cycle voltage data of the lithium ion battery without fault and the discharge cycle voltage data of the lithium ion battery with open circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time normalization (DTW) to obtain the safety threshold of the open circuit fault;

the overheating threshold calculation unit is used for inputting the temperature data of the lithium ion battery charging and discharging cycle probe which does not have a fault and the temperature data of the lithium ion battery charging and discharging cycle probe which has an overheating fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain a safety threshold of the overheating fault;

and the supercooling threshold value calculating unit is used for inputting the temperature data of the lithium ion battery charge-discharge cycle probe which does not have a fault and the temperature data of the lithium ion battery charge-discharge cycle probe which has the supercooling fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the signal noise reduction model and the characteristic extraction model, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain the safety threshold value of the supercooling fault.

6. The lithium ion battery fault diagnosis method according to claim 1, wherein the lithium ion battery fault diagnosis model of step S5 specifically includes:

the short-circuit fault judging unit is used for determining that a short-circuit fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the signal noise reduction model and the characteristic extraction model of the lithium ion battery charging circulating voltage data and a charging voltage characteristic datum line reaches a short-circuit threshold value; the charging voltage characteristic reference line refers to a characteristic parameter curve obtained after the charging voltage data of the battery for the first time is input into a signal noise reduction model and a characteristic extraction model;

the circuit breaking fault determination unit is used for determining that circuit breaking fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the discharge cycle voltage data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and a discharge voltage characteristic datum line reaches an overdischarge threshold value; the discharge voltage characteristic reference line refers to a characteristic parameter curve obtained after the first discharge voltage data of the battery is input into a signal noise reduction model and a characteristic extraction model;

the overheating fault determination unit is used for determining that overheating faults occur when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging cycle probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches an overheating threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model;

the supercooling fault judging unit is used for determining that a supercooling fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging circulation probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches a supercooling threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model.

7. The lithium ion battery fault diagnosis method according to claim 1, wherein step S6 specifically includes:

acquiring charge-discharge cycle voltage and probe temperature data in the use process of a battery to be tested, inputting the data into a signal noise reduction model and a characteristic extraction model, and obtaining a charge voltage characteristic parameter curve, a charge probe temperature characteristic parameter curve, a discharge voltage characteristic parameter curve and a discharge probe temperature characteristic parameter curve;

comparing the charging voltage characteristic parameter curve with the charging voltage characteristic reference line to obtain the similarity degree of the charging voltage curve;

comparing the charging probe temperature characteristic parameter curve with the charging temperature characteristic datum line to obtain the similarity degree of the charging temperature curve;

comparing the discharge voltage characteristic parameter curve with the discharge voltage characteristic datum line to obtain the similarity degree of the discharge voltage curve;

and comparing the characteristic parameter curve of the discharge probe temperature with the discharge temperature characteristic datum line to obtain the similarity degree of the discharge temperature curve.

8. The lithium ion battery fault diagnosis method according to claim 1, wherein the similarity degree is input into a lithium ion battery fault diagnosis model, and if a threshold is reached, a corresponding fault type alarm signal is sent, specifically comprising:

when the similarity degree of the charging voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the charging voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the charging temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

when the similarity degree of the charging temperature curves reaches a safety threshold of the overheating fault, determining that the overheating fault occurs, and sending an overheating alarm signal;

when the similarity degree of the discharge voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the discharge voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the discharge temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

and when the similarity degree of the discharge temperature curve reaches the safety threshold of the overheating fault, determining that the overheating fault occurs, and sending an overheating alarm signal.

Technical Field

The invention belongs to the field of battery fault diagnosis, and particularly relates to a lithium ion battery fault diagnosis method.

Background

In recent years, secondary lithium ion batteries have been widely used in the fields of C products, electric vehicles, energy storage, and the like, due to their high energy density, long service life, low self-discharge rate, no memory effect, and the like. Particularly, with the increasingly prominent environmental protection problem, the use of lithium ion batteries in the field of electric automobiles is in an almost linear increasing trend. However, due to the instability, abuse and technological development of lithium ion batteries, the technical requirements of battery slimness and high energy density are more severe, and frequent safety accidents are attracting more and more attention. Therefore, the invention is urgent for a method capable of detecting the power battery fault in real time and predicting the result accurately.

The lithium ion battery is a complex nonlinear system which changes in real time, the internal chemical reaction mechanism is complex, the external performance is influenced by the change of various parameters, the establishment of an electrochemical mechanism model for prediction is very complex, and the real-time prediction is difficult to achieve. In recent years, a neural network is widely applied to power battery fault diagnosis, but the method has the defects of complex structure, huge calculation amount, poor interpretability and the like, is purely driven based on data, does not consider the relation between internal microscopic chemical reaction and external macroscopic expression when a fault occurs, and has poor accuracy of a prediction result. The method provided by the invention is based on a synchronous compression continuous wavelet transform noise reduction and characteristic parameter extraction method, the extracted characteristic parameters can represent the change of chemical reaction energy with different frequencies in a lithium ion battery when a fault occurs, the external macroscopic expression is associated with the internal microscopic change, and the accurate, efficient and real-time power battery fault diagnosis can be realized.

Disclosure of Invention

In order to overcome the defects of the method, the method for diagnosing the lithium ion battery faults is provided, modeling is not needed, the calculated amount is small, and real-time online detection can be achieved; and the external macroscopic expression is related to the frequency of the internal chemical reaction, so that the prediction accuracy is greatly improved.

The invention is realized by at least one of the following technical schemes.

A lithium ion battery fault diagnosis method comprises the following steps:

s1, acquiring battery data when the lithium ion battery fails and various types of faults occur;

s2, carrying out noise reduction processing on the data obtained in the step S1 by using a signal noise reduction model;

s3, calculating characteristic parameters representing chemical reactions of different frequencies in the lithium ion battery through a characteristic extraction model;

s4, calculating the safety threshold of the lithium ion battery by comparing the similarity of the characteristic parameter curves of the lithium ion battery without faults and the lithium ion battery with various types of faults;

s5, determining the corresponding fault type during alarming according to the safety threshold, and establishing a lithium ion battery fault diagnosis model;

s6, acquiring charge and discharge cycle battery data in the use process of the lithium ion battery to be diagnosed, denoising the data and calculating to obtain a characteristic parameter curve, and comparing the similarity of the characteristic parameter curve and a reference parameter curve to obtain the similarity;

and S7, inputting the similarity degree into the lithium ion battery fault diagnosis model in the step S5, and if the similarity degree reaches a threshold value, sending out a corresponding fault type alarm signal.

Further, the battery data acquired in step S1 includes the battery voltage, the probe temperature, and the corresponding fault type in the lithium ion battery history data.

Further, the noise reduction processing of the signal noise reduction model in step S2 includes the following steps:

converting the obtained lithium ion battery data from the time series s (t) into a time-frequency domain, i.e. Continuous Wavelet Transform (CWT) coefficients, using a Continuous Wavelet Transform (CWT), the obtained time-frequency domain being:

where α denotes the continuous wavelet analysis scale, τ denotes the continuous wavelet analysis time transfer window size, t denotes time, denotes the complex conjugate,<s,ψ_α,τ>denotes s (t) andthe inner product of (a) is,representing an analysis function in a continuous wavelet transform, i.e. a parent wave;

the resulting time-frequency domain is divided into a high-energy low-frequency part and a high-energy high-frequency part:

wherein, W_s(α, τ) is the resulting time-frequency domain, n_aFor the number of scales, CF (tau) is the superposition amplitude of the CWT coefficient calculated by using all continuous wavelet analysis scales a, the distribution of the CWT coefficient along the scale axis is obtained by the CF coefficient, because the existence of the low-frequency characteristic enables the obtained distribution to have two different peak values, and the two different peak values are divided by setting an optimal threshold value, so that a high-energy low-frequency part and a high-energy high-frequency part are obtained;

synchronously compressing the low-frequency part and the high-frequency part by using synchronous compression continuous wavelet transform (SS-CWT) respectively to obtain corresponding instant frequency, namely a synchronous compression continuous wavelet transform coefficient (SS-CWT) coefficient, wherein the instant frequency is as follows:

i represents a complex number, representing a partial derivative;

for the instantaneous frequency obtained by synchronous compression of the low-frequency part and the instantaneous frequency obtained by synchronous compression of the high-frequency part, different methods are adopted for noise reduction, and the method specifically comprises the following steps:

for a low-frequency part, introducing a soft interval screening characteristic to filter noise, wherein the soft interval is as follows:

lambda is the set threshold value and is the threshold value,indicates the characteristics after screening, omega_sRepresenting continuous wavelet transform coefficients;

for the high frequency part, calculating the superposed amplitude CF of the front signal segment, screening the characteristics of the high frequency part by using a hard interval, and filtering the main noise, wherein the hard interval is as follows:

wherein λ is_nTo set the threshold value, M_max＝mean(max|T_nI), and T)_rRespectively calculating SS-CWT coefficients of narrow frequency bands corresponding to the two peak values obtained after CF calculation;

combining the SS-CWT coefficient of the high-frequency part and the SS-CWT coefficient of the low-frequency part after noise reduction to form a time-frequency domain after noise reduction, and inverting to a time sequence signal;

and (3) converting the denoised time sequence signal again through continuous wavelet transform to obtain a Continuous Wavelet Transform (CWT) coefficient, and performing post-denoising again by using a CT threshold, wherein the CT threshold is as follows:

where λ is the set threshold, 0<γ<Lambda is more than or equal to 0 and less than or equal to α and less than or equal to 1, gamma is a truncation value, when the continuous wavelet transform coefficient is less than the truncation value, the coefficient is set to 0,for the selected characteristics, sgn (W)_s) Indicating that if the continuous wavelet transform coefficient is positive, the output is 1, otherwise-1, W is output_sRepresenting continuous wavelet transform coefficients;

and the signal denoising model outputs a Continuous Wavelet Transform (CWT) coefficient obtained by the post-denoising.

Further, step S3 specifically includes:

calculating a parameter DF by multi-scale envelope superposition of Continuous Wavelet Transform (CWT) coefficients output by a signal denoising module, wherein DF is:

where n is the number of scales, E (α, tau) is the envelope function of the Continuous Wavelet Transform (CWT) coefficients, and the formula isHere, theIs the result of the hilbert transform of the continuous wavelet transform coefficients;

calculating an operating energy ratio ER using the DF obtained above₁Said operating energy ratio ER₁Comprises the following steps:

wherein L is the length of the energy collection window before and after the time transfer window size τ;

using ER obtained as described above₁Calculating characteristic parameter ER for representing chemical reactions with different frequencies in lithium ion battery₂The ER₂Comprises the following steps:

ER₂(τ)＝ER₁(τ)|DF(α)|

the characteristic extraction model outputs the characteristic parameter ER obtained above₂。

Further, step S4 is to calculate a safety threshold of the lithium ion battery, and specifically includes:

the short-circuit threshold calculation unit is used for inputting the charging cycle voltage data of the lithium ion battery without fault and the charging cycle voltage data of the lithium ion battery with short-circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the charging cycle voltage data and the charging cycle voltage data to obtain a safety threshold of the short-circuit fault;

the open circuit threshold calculation unit is used for inputting the discharge cycle voltage data of the lithium ion battery without fault and the discharge cycle voltage data of the lithium ion battery with open circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time normalization (DTW) to obtain the safety threshold of the open circuit fault;

the overheating threshold calculation unit is used for inputting the temperature data of the lithium ion battery charging and discharging cycle probe which does not have a fault and the temperature data of the lithium ion battery charging and discharging cycle probe which has an overheating fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain a safety threshold of the overheating fault;

and the supercooling threshold value calculating unit is used for inputting the temperature data of the lithium ion battery charge-discharge cycle probe which does not have a fault and the temperature data of the lithium ion battery charge-discharge cycle probe which has the supercooling fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the signal noise reduction model and the characteristic extraction model, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain the safety threshold value of the supercooling fault.

Further, the lithium ion battery fault diagnosis model in step S5 specifically includes:

the short-circuit fault judging unit is used for determining that a short-circuit fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the signal noise reduction model and the characteristic extraction model of the lithium ion battery charging circulating voltage data and a charging voltage characteristic datum line reaches a short-circuit threshold value; the charging voltage characteristic reference line refers to a characteristic parameter curve obtained after the charging voltage data of the battery for the first time is input into a signal noise reduction model and a characteristic extraction model;

the circuit breaking fault determination unit is used for determining that circuit breaking fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the discharge cycle voltage data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and a discharge voltage characteristic datum line reaches an overdischarge threshold value; the discharge voltage characteristic reference line refers to a characteristic parameter curve obtained after the first discharge voltage data of the battery is input into a signal noise reduction model and a characteristic extraction model;

the overheating fault determination unit is used for determining that overheating faults occur when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging cycle probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches an overheating threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model;

the supercooling fault judging unit is used for determining that a supercooling fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging circulation probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches a supercooling threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model.

Further, step S6 specifically includes:

acquiring charge-discharge cycle voltage and probe temperature data in the use process of a battery to be tested, inputting the data into a signal noise reduction model and a characteristic extraction model, and obtaining a charge voltage characteristic parameter curve, a charge probe temperature characteristic parameter curve, a discharge voltage characteristic parameter curve and a discharge probe temperature characteristic parameter curve;

comparing the charging voltage characteristic parameter curve with the charging voltage characteristic reference line to obtain the similarity degree of the charging voltage curve;

comparing the charging probe temperature characteristic parameter curve with the charging temperature characteristic datum line to obtain the similarity degree of the charging temperature curve;

comparing the discharge voltage characteristic parameter curve with the discharge voltage characteristic datum line to obtain the similarity degree of the discharge voltage curve;

and comparing the characteristic parameter curve of the discharge probe temperature with the discharge temperature characteristic datum line to obtain the similarity degree of the discharge temperature curve.

Further, inputting the similarity degree into a lithium ion battery fault diagnosis model, and if the similarity degree reaches a threshold value, sending out a corresponding fault type alarm signal, which specifically comprises the following steps:

when the similarity degree of the charging voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the charging voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the charging temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

when the similarity degree of the charging temperature curves reaches a safety threshold of the overheating fault, determining that the overheating fault occurs, and sending an overheating alarm signal;

when the similarity degree of the discharge voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the discharge voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the discharge temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

and when the similarity degree of the discharge temperature curve reaches the safety threshold of the overheating fault, determining that the overheating fault occurs, and sending an overheating alarm signal.

Compared with the prior art, the invention has the following beneficial effects: the method provided by the invention is based on a synchronous compression continuous wavelet transform noise reduction and characteristic parameter extraction method, the extracted characteristic parameters can represent the change of chemical reaction energy with different frequencies in a lithium ion battery when a fault occurs, the external macroscopic expression is associated with the internal microscopic change, and the accurate, efficient and real-time power battery fault diagnosis can be realized.

Drawings

Fig. 1 is a flow chart of a lithium ion battery fault diagnosis method and system according to an embodiment of the present invention;

FIG. 2 is a flowchart of a signal noise reduction model according to an embodiment of the present invention;

FIG. 3 is a flowchart of the feature parameter extraction model according to an embodiment of the present invention;

fig. 4 is a flowchart of a prediction process in the lithium ion battery fault diagnosis method according to the embodiment of the present invention;

fig. 5 is a first charging cycle voltage curve of a failed lithium ion battery according to an embodiment of the present invention and a charging cycle voltage curve when a short-circuit fault occurs;

fig. 6 is a first charging cycle voltage characteristic curve of a failed lithium ion battery and a charging cycle voltage characteristic curve when a short-circuit fault occurs according to an embodiment of the present invention.

Detailed description of the invention

The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

The invention aims to provide a lithium ion battery online fault diagnosis method and a lithium ion battery online fault diagnosis system, which are used for realizing online real-time fault detection of a lithium ion battery.

In order to make the aforementioned objects and features of the present invention more comprehensible, the present invention will be described in further detail with reference to the accompanying drawings and embodiments.

The lithium ion battery fault diagnosis method shown in fig. 1 includes the following steps:

step 1, acquiring battery data when the lithium battery fails and various types of faults occur, and specifically comprising the following steps:

and extracting the battery voltage, the probe temperature and the corresponding fault type in the lithium ion battery historical data.

Step 2, carrying out noise reduction processing on the data obtained in the step S1 by using a signal noise reduction model; fig. 2 is a flowchart of a noise reduction model according to an embodiment, which specifically includes the following steps:

step 201, using Continuous Wavelet Transform (CWT) to convert the obtained lithium ion battery data from time series to time-frequency domain, i.e. Continuous Wavelet Transform (CWT) coefficients, where the obtained time-frequency domain is:

where α denotes the continuous wavelet analysis scale, τ denotes the continuous wavelet analysis time transfer window size, denotes the complex conjugate, t denotes time,<s,ψ_α,τ>denotes s (t) andphi (t) represents an analytical function in a continuous wavelet transform, i.e. the mother wave;

step 202, dividing the obtained time-frequency domain into a high-energy low-frequency part and a high-energy high-frequency part;

wherein, W_s(α, τ) is the resulting time-frequency domain, n_aFor the number of scales, CF (τ) is the superimposed amplitude of the CWT coefficients calculated using all the continuous wavelet analysis scales α, the distribution of the CWT coefficients along the scale axis is derived from the CF coefficients because the presence of the low frequency feature causes the resulting distribution to have two different peaks, which are divided by setting an optimal threshold, thereby obtaining the high energy low frequency part and the high energy high frequency part.

Step 203, synchronously compressing the obtained high frequency part and low frequency part by using synchronous compression continuous wavelet transform (SS-CWT) respectively to obtain an instant frequency, namely a synchronous compression continuous wavelet transform coefficient (SS-CWT) coefficient, wherein the instant frequency is:

i represents a complex number, representing a partial derivative;

step 204, for the instantaneous frequency obtained by the synchronous compression of the low frequency part and the instantaneous frequency obtained by the synchronous compression of the high frequency part, different methods are adopted to reduce noise, specifically as follows:

for a low-frequency part, introducing a soft interval screening characteristic to filter noise, wherein the soft interval is as follows:

lambda is the set threshold value and is the threshold value,indicates the characteristics after screening, omega_sRepresenting continuous wavelet transform coefficients;

for the high frequency part, calculating the superposed amplitude CF of the front signal segment, screening the characteristics of the high frequency part by using a hard interval, and filtering the main noise, wherein the hard interval is as follows:

wherein λ is_nTo set the threshold value, M_max＝mean(max|T_n|),T_n，T_rRespectively calculating SS-CWT coefficients of narrow frequency bands corresponding to the two peak values obtained after CF calculation;

step 205, combining the SS-CWT coefficient of the high frequency part and the SS-CWT coefficient of the low frequency part after noise reduction to form a time-frequency domain after noise reduction, and then reversely converting the SS-CWT coefficient and the SS-CWT coefficient into a time sequence signal after noise reduction by using the reverse sequence of the steps;

and step 206, converting the denoised time sequence signal again through continuous wavelet transform to obtain a Continuous Wavelet Transform (CWT) coefficient, and performing post-denoising again by using a CT threshold, wherein the CT threshold is as follows:

where λ is the set threshold, 0<γ<Lambda is more than or equal to 0 and less than or equal to α and less than or equal to 1, gamma is a truncated value, when the continuous wavelet transform coefficient is less than the truncated value, the coefficient is set to 0,for the selected characteristics, sgn (W)_s) It means that if the continuous wavelet transform coefficient is positive, the output is 1, otherwise-1 is output. W_sRepresenting continuous wavelet transform coefficients.

And step 207, outputting the Continuous Wavelet Transform (CWT) coefficient obtained by the post-denoising.

Step 3, calculating through a feature extraction model to obtain feature parameters representing chemical reactions at different frequencies inside the lithium ion battery, where fig. 3 is a work flow diagram of the feature extraction model of this embodiment, and specifically includes:

step 301, calculating a parameter DF by multi-scale envelope superposition of Continuous Wavelet Transform (CWT) coefficients output by the signal denoising model of step S2, where DF is:

in the formula, n_aIs the number of scales, E (α, tau) is the envelope function of the Continuous Wavelet Transform (CWT) coefficients, and the calculation formula isHere, theIs the result of the hilbert transform of the continuous wavelet transform coefficients, n being the number of scales.

Step 302, calculating the running energy ratio ER using the DF obtained above₁Said operating energy ratio ER₁Comprises the following steps:

wherein L is the length of the energy collection window before and after the time transfer window size τ;

step 303, Using ER obtained above₁Calculating characteristic parameter ER for representing chemical reactions with different frequencies in lithium ion battery₂The ER₂Comprises the following steps:

ER₂(τ)＝ER₁(τ)|DF(α)|

step 304, outputting the characteristic parameter ER obtained above₂。

Step 4, calculating the safety threshold of the lithium ion battery by comparing the similarity of the characteristic parameter curves of the lithium ion battery without faults and the lithium ion battery with various types of faults, and specifically comprises the following steps:

the circuit threshold value calculation unit is used for inputting the charging cycle voltage data of the lithium ion battery without fault and the charging cycle voltage data of the lithium ion battery with short circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time normalization (DTW) to obtain the safety threshold value of the short circuit fault;

the open circuit threshold calculation unit is used for inputting the discharge cycle voltage data of the lithium ion battery without fault and the discharge cycle voltage data of the lithium ion battery with open circuit fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time normalization (DTW) to obtain the safety threshold of the open circuit fault;

the overheating threshold calculation unit is used for inputting the temperature data of the lithium ion battery charging and discharging cycle probe which does not have a fault and the temperature data of the lithium ion battery charging and discharging cycle probe which has an overheating fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the two models, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain a safety threshold of the overheating fault;

the supercooling threshold value calculating unit is used for inputting the temperature data of the lithium ion battery charge-discharge cycle probe which does not have a fault and the temperature data of the lithium ion battery charge-discharge cycle probe which has a supercooling fault into the signal noise reduction model and the characteristic extraction model to obtain characteristic parameter curves corresponding to the signal noise reduction model and the characteristic extraction model, and comparing the similarity degree of the two characteristic parameter curves by using dynamic time regression (DTW) to obtain a safety threshold value of the supercooling fault;

step 5, determining the corresponding fault type during alarming according to the safety threshold, and establishing a lithium ion battery fault diagnosis model, which specifically comprises the following steps:

the short-circuit fault judging unit is used for determining that a short-circuit fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the signal noise reduction model and the characteristic extraction model of the lithium ion battery charging circulating voltage data and a charging voltage characteristic datum line reaches a short-circuit threshold value; the charging voltage characteristic reference line refers to a characteristic parameter curve obtained after the charging voltage data of the battery for the first time is input into a signal noise reduction model and a characteristic extraction model;

the circuit breaking fault determination unit is used for determining that circuit breaking fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the discharge cycle voltage data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and a discharge voltage characteristic datum line reaches an overdischarge threshold value; the discharge voltage characteristic reference line refers to a characteristic parameter curve obtained after the first discharge voltage data of the battery is input into a signal noise reduction model and a characteristic extraction model;

the overheating fault determination unit is used for determining that overheating faults occur when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging cycle probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches an overheating threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model;

the supercooling fault judging unit is used for determining that a supercooling fault occurs when the similarity degree of a characteristic parameter curve obtained by inputting the charging and discharging circulation probe temperature data of the lithium ion battery into the signal noise reduction model and the characteristic extraction model and the charging and discharging temperature characteristic datum line reaches a supercooling threshold value; the charge and discharge temperature characteristic datum line refers to a characteristic parameter curve obtained after the charge and discharge temperature data of the battery for the first time are input into a signal noise reduction model and a characteristic extraction model.

Step 6, acquiring charge and discharge cycle battery data in the use process of the lithium ion battery to be diagnosed, denoising the data and calculating to obtain a characteristic parameter curve, and comparing the similarity of the characteristic parameter curve and a reference parameter curve to obtain the similarity degree, wherein the method specifically comprises the following steps:

acquiring charge-discharge cycle voltage and probe temperature data in the use process of a battery to be tested, inputting the data into a signal noise reduction model and a characteristic extraction model, and obtaining a charge voltage characteristic parameter curve, a charge probe temperature characteristic parameter curve, a discharge voltage characteristic parameter curve and a discharge probe temperature characteristic parameter curve;

comparing the charging voltage characteristic parameter curve with the charging voltage characteristic datum line by using dynamic time regression (DTW) to obtain the similarity degree of the charging voltage curve;

comparing the charging probe temperature characteristic parameter curve with the charging temperature characteristic datum line by using dynamic time regression (DTW) to obtain the similarity degree of the charging temperature curve;

comparing the discharge voltage characteristic parameter curve with the discharge voltage characteristic datum line by using Dynamic Time Warping (DTW) to obtain the similarity degree of the discharge voltage curve;

and comparing the characteristic parameter curve of the discharge probe temperature with the discharge temperature characteristic datum line by using Dynamic Time Warping (DTW) to obtain the similarity degree of the discharge temperature curve.

And 7, inputting the similarity degree into the lithium ion battery fault diagnosis model, and if the similarity degree reaches a threshold value, sending out a corresponding fault type alarm signal, wherein the method specifically comprises the following steps:

when the similarity degree of the charging voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the charging voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the charging temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

when the similarity degree of the charging temperature curves reaches a safety threshold of the overheating fault, determining that the overheating fault occurs, and sending an overheating alarm signal;

when the similarity degree of the discharge voltage curves reaches a safety threshold value of the short-circuit fault, determining that the short-circuit fault occurs, and sending a short-circuit alarm signal;

when the similarity degree of the discharge voltage curves reaches the safety threshold of the open circuit fault, determining that the open circuit fault occurs, and sending an open circuit alarm signal;

when the similarity degree of the discharge temperature curves reaches the safety threshold of the supercooling fault, determining that the supercooling fault occurs, and sending a supercooling alarm signal;

when the similarity degree of the discharge temperature curves reaches the safety threshold of the overheating fault, the overheating fault is determined to occur, and an overheating alarm signal is sent out

Fig. 4 is a flowchart of a prediction process in the lithium ion battery fault diagnosis method according to the embodiment of the present invention, which specifically includes:

inputting the obtained similarity into a lithium ion battery fault diagnosis model, firstly judging whether the similarity reaches a threshold value, and if not, not sending an alarm signal; if the threshold value is reached, sending an alarm signal, judging whether the threshold value reaches an overheating fault threshold value, if so, sending an overheating alarm signal, and if not, entering the next step; judging whether a supercooling fault threshold value is reached, if so, sending a supercooling alarm signal, and if not, entering the next step; judging whether a short circuit threshold value is reached, if so, sending a short circuit alarm signal, and if not, entering the next step; and judging whether the threshold value of the open circuit is reached, and if so, sending an open circuit alarm signal.