CN111967190B - Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition - Google Patents
Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition Download PDFInfo
- Publication number
- CN111967190B CN111967190B CN202010857310.0A CN202010857310A CN111967190B CN 111967190 B CN111967190 B CN 111967190B CN 202010857310 A CN202010857310 A CN 202010857310A CN 111967190 B CN111967190 B CN 111967190B
- Authority
- CN
- China
- Prior art keywords
- lithium
- safety
- battery
- lithium dendrite
- percentage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
- H01M10/00—Secondary cells; Manufacture thereof
- H01M10/42—Methods or arrangements for servicing or maintenance of secondary cells or secondary half-cells
- H01M10/4285—Testing apparatus
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Manufacturing & Machinery (AREA)
- Chemical & Material Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Electrochemistry (AREA)
- General Chemical & Material Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Secondary Cells (AREA)
Abstract
A lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition belong to the technical field of power battery safety degree evaluation. The invention aims to solve the problem that the safety of a power battery cannot be quantitatively expressed and evaluated in the prior art. The safety state of the lithium battery to be evaluated is judged by acquiring the appearance image of the lithium dendrite to train a percentage quantification model of the CNN of the lithium dendrite. Firstly, collecting lithium dendrite images of a lithium battery, classifying and processing the collected lithium dendrite morphology images, secondly, sending the processed lithium dendrite morphology images into an established percentage quantization model of lithium dendrite CNN for training, establishing a mathematical model for quantizing the percentage of safety degree of the lithium battery and a classification model of safety grade of the lithium battery, and finally, calculating and evaluating the percentage of safety degree and the safety grade of the lithium battery to be detected by using the trained percentage quantization model of the lithium dendrite CNN; the invention solves the problem that the prior art can not carry out quantitative evaluation on the safety of the battery.
Description
Technical Field
The invention relates to the field of battery safety degree evaluation, in particular to a lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition.
Background
With the increasing pace of commercialization of electric vehicles in the global market, the demand for high-power and high-energy power batteries is rapidly increasing, and the high-power and high-energy power batteries also expose numerous safety problems, so that the safety of the batteries is receiving more and more attention. In particular, in recent years, lithium batteries have been widely used in various fields related to battery applications as a representative of high power and high energy batteries, but accidents such as spontaneous combustion and explosion of lithium batteries occur in the news, and the safety of lithium batteries is gradually valued by manufacturers and users. At present, lithium batteries in China are still in an initial stage in the technical research and development level, and have a plurality of problems in the safety aspect of the lithium batteries in the application field of the lithium batteries.
The lithium ion battery is a complex electrochemical system, the failure mechanism of the lithium battery is complex, and the failure mode of the lithium battery is influenced by a plurality of factors, such as the ambient temperature, the discharge depth, the charge and discharge current and the like. Although the state parameters of the lithium battery, such as voltage, current, temperature and the like, can be measured in real time, and the parameters of internal resistance, capacity, SOC and the like can also be obtained through calculation through actually measured parameters, the safety of the lithium battery cannot be measured, and is a variable which is influenced by multiple factors and changes at any time, and the guarantee of the safety is also a precondition for normal application of a lithium battery system. The problem of lithium battery safety quantification also becomes a key point and a difficulty of current battery application research and safety research. At present, domestic scholars are few in the aspect of quantitative research on lithium battery safety, and mainly focus on the aspect of a lithium battery fault diagnosis method. At present, fault diagnosis technology in the lithium battery industry is relatively mature, when a lithium battery breaks down, the fault reason of the lithium battery can be accurately found, but the fault diagnosis is only after the lithium battery breaks down, a lithium battery tester judges the fault problem, and the fault diagnosis cannot prevent the lithium battery from breaking down. In fact, the formation of the lithium battery fault is a gradually changing process, for example, the lithium battery can be subjected to safety evaluation in the using process, and quantitative indexes are given, so that the lithium battery fault diagnosis method has an important role in preventing lithium battery accidents and guaranteeing the life safety of users. In addition, the performance indexes of the lithium battery comprise capacity, energy density, charge-discharge rate, voltage, service life, internal resistance, self-discharge, working temperature and the like. Although the parameters can reflect the use condition of the lithium battery, the danger of the lithium battery cannot be accurately reflected by the abnormity of any parameter. Therefore, the parameters can not give a user of the lithium battery the intuitive feeling of the safety of the power battery, the parameters of the lithium battery are many and dense, and the slight parameter change can also have huge potential safety hazards. Therefore, in practical application, the lithium battery may still work normally, but a large potential safety hazard exists at this time, how to judge the safety state of the battery in detection through a simple battery test is not only achieved, and how to prompt a user that the safety of the lithium battery is a technical problem to be solved urgently in the battery industry. The problem is not only limited to lithium batteries, but also no method for quantifying the safety degree of various batteries exists in any current battery.
Disclosure of Invention
In order to solve the problems, the invention provides a lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition according to the probability of key faults occurring in the use process of a lithium ion power battery, and solves the problem that the safety of the battery cannot be quantitatively evaluated in the prior art.
The invention provides a lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition, which comprises the following steps of:
acquiring morphology images of the lithium dendrites of thousands of lithium batteries to be evaluated, and preprocessing the images;
sending the processed lithium dendrite morphology image into an established percentage quantification model of the lithium dendrite CNN for training;
the percentage quantification model of a large number of trained lithium dendrite CNN is used for realizing the accurate quantification of the safety percentage of the lithium battery, namely, the percentage accurate value of the safety of the lithium battery is calculated;
establishing a safety degree comparison table, wherein the safety degree comparison table is composed of a plurality of safety intervals, and the safety intervals correspond to the battery safety conditions at the current moment; and matching the safety degree value obtained by the estimation module with the safety interval to obtain the battery safety condition at the current moment.
The percentage numerical value of the safety degree of the lithium battery is divided into five intervals, the five intervals of the percentage numerical value of the safety degree of the lithium battery are set to five safety levels of extreme safety, good safety, general safety, potential danger and serious danger corresponding to the safety degree level, and the safety degree level to which the percentage quantitative numerical value of the safety degree belongs is judged;
finally, calculating and evaluating the safety percentage and the safety grade of the lithium battery to be detected by using the trained percentage quantization model of the lithium dendrite CNN, and outputting the safety percentage numerical value and the safety grade of the lithium battery to be detected;
further, the pretreatment process comprises:
acquiring lithium dendrite morphology images of a lithium battery, and classifying thousands of lithium dendrite morphology images;
data balance is carried out on the lithium dendrite image of the lithium battery, and the problem that the obtained lithium dendrite image has data unbalance is solved;
and carrying out numerical conversion on the lithium dendrite morphology graph obtained through data balance to obtain a lithium battery safety degree percentage quantification data set.
Further, the classification of the lithium dendrite morphology image is as follows: a no lithium ion deposition topography image, a no bifurcation topography image, a cluster-like topography image, a visible distinct bifurcation structure topography image, and a visible lithium dendrite has pierced the separator topography image.
Further, the percentage quantization model of the lithium dendrite CNN includes 16 layers, and the specific steps from input to output are: the first time is an input layer, the second layer is subjected to twice convolution of 64 convolution kernels, a pooling layer is performed for the first time, the second time is subjected to 128 convolution kernel convolution layers, the first pooling layer is performed, 512 convolution kernel convolutions are repeated twice, and the second layer sequentially passes through the pooling layer and a three-time full-connection layer.
Further, the percentage quantification value of the safety degree of the lithium battery is obtained by the following formula:
wherein, w j The weight coefficient which influences the battery safety and corresponds to the extracted jth characteristic value is adopted, n is the total number of the characteristics, f j Is the characteristic value of the j-th lithium dendrite characteristic extracted.
Further, the characteristic value of the jth lithium dendrite characteristic is:
wherein I i Is the characteristic value of the ith lithium dendrite characteristic in the image.
The invention provides a lithium battery safety degree evaluation device based on lithium dendrite morphology image recognition, which comprises the following steps:
the estimation module is used for estimating the safety degree of the current state of the battery according to the lithium battery safety degree estimation method based on the lithium dendrite morphology image recognition in the embodiment;
the interval matching module is used for establishing a safety degree comparison table, the safety degree comparison table is composed of a plurality of safety intervals, and the safety intervals correspond to the battery safety conditions at the current moment; matching the safety degree value obtained by the estimation module with the safety interval to obtain the battery safety condition at the current moment;
and the display module is used for displaying the safety degree value.
As described above, the lithium battery safety degree evaluation method for lithium dendrite morphology image recognition based on the convolutional neural network provided by the invention has the following effects:
1. according to the method, the characteristic extraction is carried out on the lithium dendrite morphology image of the lithium battery through the CNN model, the safety degree of the lithium battery to be overhauled is quantized, the percentage numerical value of the safety degree of the lithium battery to be specifically overhauled is output, and the safety grade classification of the lithium battery to be overhauled is realized according to the division of the quantized numerical value interval.
2. The method has the advantages of simple training model, good sample adaptability, convenient updating of the calculation result and suitability for the calculation of the safety degree of the lithium ion battery.
3. The lithium battery safety degree evaluation method aiming at the image recognition of the lithium dendrite morphology based on the convolutional neural network is adopted, the strong image processing training capacity of the method is linked with the lithium battery lithium dendrite morphology image, and the morphology of the lithium battery lithium dendrite represents the safety state of the lithium battery, so the lithium battery safety degree evaluation method aiming at the image recognition of the lithium dendrite morphology based on the convolutional neural network has unique advantages.
4. The method is suitable for calculating the safety degree of various batteries, and has wide applicability, easy realization of a convolution network and more application occasions.
In conclusion, the lithium battery safety degree evaluation method aiming at the lithium dendrite morphology image recognition based on the convolutional neural network is very suitable for safety degree evaluation of various batteries and has practicability.
Drawings
FIG. 1 is a flowchart of a method for evaluating safety of a lithium battery based on convolution according to an embodiment of the present invention;
FIG. 2 is a convolutional neural network model for lithium dendrite morphology identification according to an embodiment of the present invention;
FIG. 3 is a diagram of a layer of convolutional neural network architecture in accordance with an embodiment of the present invention;
FIG. 4 is a schematic block diagram of a method for calculating a safety level of a lithium battery according to a weight coefficient in an embodiment of the present invention;
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the features in the following embodiments and examples may be combined with each other without conflict.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention, and the drawings only show the components related to the present invention rather than the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
The safety of the battery refers to that the battery does not burn, explode, generate toxic and harmful gases and do not harm a user in the using process, and the quantitative description of the safety degree of the battery in the using process is called the safety degree of the battery. The safety degree of the power battery is an important index for evaluating the safety degree of the battery and is an important parameter for describing the safety degree of the battery. Because the test conditions of the battery safety degree under different voltages, different currents and different temperatures can show different results, the deviation needs to be classified and quantified, and the technical problem which is not solved by the industry is solved.
In the prior art, research related to battery safety mainly develops around fault diagnosis technologies, and the methods are directed at diagnosing the cause of the battery fault after the battery fault occurs, and improving the battery according to the diagnosis result, and the method cannot substantially prevent the battery fault from occurring, but during the use process of the battery, the fault behavior of the battery is a gradually changing process, and in many cases, the battery reaches an extreme aging condition of combustion and explosion when the fault behavior of the battery is not obviously shown, so that great loss is caused to personnel and property.
In the using process of the lithium battery, the staged safety of the lithium battery can be stably embodied and quantified, namely the lithium battery can locally generate a temperature hot spot due to the enhancement of the surface exchange current density, the temperature hot spot can initiate a remarkable growth phenomenon of lithium dendrite around the temperature hot spot, the lithium dendrite is lithium whisker growing in the battery, and the growth of the lithium dendrite can cause internal short circuit. More importantly, the internal short circuit of the lithium battery can further cause the local temperature to rise, so that the experiment finds the two-way relationship between the growth of the lithium dendrite and the local temperature rise: the temperature hot spots can cause remarkable lithium metal growth, further cause internal short circuit of the battery, further increase local temperature in turn, and increase the risk of thermal runaway of the battery, thereby causing danger of equipment equipped with the lithium battery. The above findings give some relationship between lithium dendrite growth and battery safety. The phasic variation of the lithium dendrite growth morphology provides the possibility to quantify the battery safety.
The embodiment provides a lithium battery safety degree evaluation method aiming at lithium dendrite morphology image recognition based on a convolutional neural network, as shown in fig. 1. The mentioned implementation steps are as follows:
the invention provides a lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition, which comprises the following steps of:
s1, obtaining a lithium dendrite morphology image X of a lithium battery to be evaluated, and preprocessing the image, wherein the collected lithium dendrite morphology image is derived from lithium dendrite morphology images obtained by the lithium battery under different external conditions, and the preprocessing process of the image comprises the following steps:
s11, acquiring morphology images of the lithium dendrites of thousands of lithium batteries, and classifying the morphology images of the lithium dendrites;
the lithium dendrites have different appearances in different battery environments and at different moments, such as bryoid lithium, filiform lithium, needle-tip lithium, whisker lithium, shrubbery lithium, dendritic lithium and the like. This phenomenon of name diversification is caused by the fact that researchers describe them differently. Can be briefly classified into 5 types: (1) no lithium ion deposition occurs, no lithium ion is converted into an atomic form, and the lithium ion on the surface of the negative electrode of the lithium battery is tiled; (2) no bifurcation, single growth, such as filiform lithium, needle-tip lithium, whisker lithium; clusters, growing similar to dough fermentation processes, such as bryozoan, shrubbery lithium; (3) clusters, growing similar to dough fermentation processes, such as bryozoan, shrubbery lithium; (4) obvious branched structures can be seen, branches are sparse, the most dangerous lithium dendritic structure is formed, and the membrane is easy to puncture, such as dendritic lithium; (5) it can be seen that the lithium dendrites have pierced through the separator, and a local short circuit test is performed on the lithium battery to find that the lithium battery has been locally short-circuited. The initial nucleation of lithium dendrites and the generation of lithium dendrites during growth can be divided into 3 stages. In the first stage, after the battery is assembled, due to the high activity of metallic lithium, a transient reaction can occur when the battery is contacted with components such as an organic solvent in an electrolyte, and an SEI film is formed, i.e., the SEI film is formed earlier than lithium dendrites are generated. The dense SEI film can prevent the electrolyte from further reacting with metallic lithium, and is a good ion conductor, but an electronic insulator. Li + may be deposited on the surface of the electrode through the SEI film, but its deposition distribution is not uniform due to the influence of lithium, electrolyte, the intrinsic characteristics of the SEI film, and charge and discharge conditions. The second phase, the nucleation phase, is the continued accumulation of uneven precipitates, causing bulges in some places until the original SEI film is broken. And finally, entering a growth stage, and continuously growing in the length direction after the original SEI film is punctured to form visible dendrites. Meanwhile, the SEI film continuously reacts and grows along with the growth of the metal lithium dendrite, but is always coated on the surface of the metal lithium. Generally, the number of lithium dendrites is mainly determined by the nucleation stage, the growth morphology of the lithium dendrites is mainly determined by the growth stage, the lithium battery image obtained by the embodiment can adopt a historical image, or a window battery is arranged on an in-situ microscope, the optical image and a laser light source are alternately shot every 40s through the same x 10 objective lens to form a nano-scale image of the lithium dendrites morphology, the window battery is a very simple platform, the in-situ microscope can be used for image collection in any laboratory, and simple electrochemical equipment can be beneficial to operation; the number of the five types of images obtained in the embodiment is respectively 400, 1000, 1500, 2000 and 400;
s12, data balance is carried out on the lithium dendrite images of the lithium battery, and the problem that the obtained lithium dendrite images are unbalanced in data is solved; the data imbalance problem is particularly common in the picture classification problem, and is represented by the fact that the proportion of one class (major class) is far greater than that of the other class (minor class), or the proportion of one response variable in a data set is far greater than that of the other class, and for the sample distribution imbalance problem, a data enhancement method can be adopted. The first data enhancement method is to randomly turn over the lithium dendrite morphology image horizontally and vertically; the second method is that the original lithium dendrite morphology image is randomly rotated for a certain angle (not more than 2 degrees), and the interpolation method used for the rotation is bicubic interpolation; the third method is to adjust the brightness and contrast of the lithium dendrite morphology image, because the change of illumination intensity can cause great influence on the imaging result, and the input picture is also denoised in the preprocessing stage.
In the embodiment, the first-class images and the fifth-class images with small quantity are subjected to left-right turning, up-down turning, 180-degree turning and the like to enable sample data of each class to be sufficient, the original images and the images subjected to the three kinds of conversion are subjected to random translation or telescopic deformation to obtain original 10-time images, namely the five-class lithium dendrite morphology images are changed into 4000, 10000, 15000, 20000 and 4000, so that the sample quantity is expanded, and the purpose is finally achieved by adopting a random undersampling mode in order to enable the five-class lithium dendrite morphology images to be close in quantity;
the first data enhancement method is to randomly turn over the lithium dendrite morphology image horizontally and vertically; the second method is that the original lithium dendrite morphology image is randomly rotated for a certain angle (not more than 2 degrees), and the interpolation method used for rotation is bicubic interpolation; the third method is to adjust the brightness and contrast of the lithium dendrite morphology image, because the variation of illumination intensity can have a great influence on the imaging result. And in the preprocessing stage, denoising the input picture.
S13, carrying out numerical conversion on the lithium dendrite morphology graph obtained through data balance to obtain a percentage quantification model data set of the lithium dendrite CNN;
converting a lithium dendrite morphology image X (300 pixel multiplied by 300 pixel) obtained through data balance into a gray image (224 pixel multiplied by 224 pixel) with a preset size, wherein the gray image is an image with each pixel having only one sampling color, and the obtained image is an image numerical matrix X * (size 224 × 224);
this embodiment is for image matrix X * Performing normalization processing to obtain a graphic matrix X * Has a pixel value range of [0,255 ]]N pixels, the image matrix X is formed as follows * The value range of the pixel(s) is scaled to the interval [0,1 ]]And the mean value is eliminated to complete the image matrix X * Is normalized to obtainNormalizing the processed image, wherein X * i Image matrix X * N is a total of n elements.
S2, establishing a percentage quantification model of the lithium dendrite CNN, and sending the preprocessed lithium dendrite morphology image into the percentage quantification model of the lithium dendrite CNN for training.
The percentage quantization model of the lithium dendrite CNN described in this embodiment includes 16 layers, and the specific steps from input to output sequentially include: the first time is input layer (gray image is 224pixel x 224 pixel), the second layer is convoluted twice by 64 convolution kernels (conv 3-64), one pooling (posing) is adopted, after two convolutions by 128 convolution kernels for the second time, one pooling (posing) is adopted again, after two convolutions by 512 convolution kernels are repeated again, pooling (posing) is carried out again, and finally, the three times of full connection are carried out, wherein the three times of convolution layers and 3 connection layers are included in total, and the structure diagram of the convolutional neural network is shown in fig. 2;
two consecutive 3 × 3 convolutions correspond to a 5 × 5 receptive field and three correspond to 7 × 7. The advantages of using three 3 × 3 convolutions instead of one 7 × 7 convolution are two-fold: 1) The inclusion of three Re Lu layers instead of one makes the decision function more discriminative; 2) Parameters are reduced, for example, input and output are all C channels, 3 convolutional layers using 3 × 3 require (3 × 3 × 3 × C) =27 × C, and 1 convolutional layer using 7 × 7 requires 7 × 7 × C =49 × C. This can be seen as applying a regularization to the 7 x 7 convolution to decompose it into 3 x 3 convolutions. The 1 × 1 convolutional layer is mainly to increase the non-linearity of the decision function without affecting the receptive field of the convolutional layer. Although the convolution operation of 1 × 1 is linear, re Lu increases the nonlinearity.
The output layer is a Soft Max classifier and the selection loss function is shown as follows:
a quantitative model parameter for representing the percentage of the CNN of the lithium dendrite that needs to be obtained by iterative training, wherein (k) The percentage quantification model representing the lithium dendrite CNN judges the corresponding model parameter when the input is the kth category; t is a unit of (t) The t-th sample representing a model of lithium dendrite CNN; 1{y n K is an indicative function, if the expression in parentheses is true, the function value of the indicative function is 1, otherwise it is 0; n is the classification number, and the invention has five categories; m is the number of training samples of the data set.
Training the percentage quantization model of the lithium dendrite CNN on the lithium dendrite image data set, wherein the training mode is small batch gradient descent, the training algorithm is a back propagation algorithm, and when a preset iteration cycle number epoch is reached, the training is finished, and the trained percentage quantization model of the lithium dendrite CNN is stored.
The safety state of the lithium battery to be evaluated is judged by utilizing a percentage quantification model of the lithium dendrite CNN, and the method specifically comprises the following steps:
s21: obtaining a lithium dendrite morphology image of a lithium battery to be evaluated, and setting the image as Y j (j＝1.2.3...)；
S22: subjecting the lithium dendrite morphology image Y j Converting all the data into 224pixel x 224pixel gray level images with preset sizes to obtain 224 x 224 numerical value matrixes, and carrying out normalization processing on the gray level images, wherein the normalization is to normalize all the data to the same range;
s23: the obtained gray level image Y after normalization processing 1 Inputting the percentage quantization model of the trained lithium dendrite CNN, extracting features, classifying the features and calculating weight coefficients of corresponding feature values, wherein the features can be the length, the morphology, the density and the like of the lithium dendrite, judging whether the lithium battery is safe or not is determined by a lithium battery safety degree formula, and if the lithium dendrite morphology image Y is extracted 1 Calculating the safety degree of the battery to be 0% through a battery safety degree formula, directly outputting the safety degree of the lithium battery to be 0%, and classifying the detected battery into an extremely dangerous range;
s3, obtaining a lithium battery safety degree percentage quantification value to which the lithium dendrite image belongs by utilizing a percentage quantification model of the lithium dendrite CNN, wherein the lithium battery safety degree percentage quantification value is obtained through the following formula:
wherein, w j A weight coefficient, w, corresponding to the extracted jth characteristic value and influencing the safety of the battery j Is obtained by training and correcting in the percentage quantization model of CNN, the weight coefficient w in this embodiment j The derivation process is similar to the conventional fuzzy control derivation process, which is not described in detail in this embodiment, and f j For the jth extracted feature value, n is the total number of features, the features are one or a combination of several of the length, morphology and concentration of the lithium dendrite, the feature selected in this embodiment is the length of the lithium dendrite, and the root mean square of the feature value of the length of the lithium dendrite is obtained by the following formula:
wherein I i Is the ith lithium dendrite length in the image.
In this embodiment, the percentage quantization model of the lithium dendrite CNN is trained on the percentage quantization dataset of the lithium dendrite CNN, the training mode is a small batch gradient descent, the training algorithm is a back propagation algorithm, when a preset iteration round number epoch is reached, the training is finished, the training mode of the multi-classification percentage quantization model of the trained lithium dendrite CNN is stored as a small batch gradient descent, the training algorithm is a back propagation algorithm, and when the preset iteration round number epoch is reached, the training is finished, and the percentage quantization model of the trained lithium dendrite CNN is stored.
S4, evaluating all remaining lithium dendrite morphology images Y of the battery safety degree for actual detection n (n = 2.3.) steps S2 to S3 are performed, and all the detection evaluation results are output.
S5, dividing the calculated lithium battery safety percentage numerical value into five intervals, setting the safety degree grades corresponding to the five lithium battery safety percentage numerical value intervals into five safety grades of extreme safety, good safety, general safety, potential danger and serious danger, and judging the safety degree grade to which the safety percentage quantitative numerical value belongs;
and S6, finally, calculating and evaluating the safety percentage and the safety grade of the lithium battery to be detected by using the trained percentage quantification model of the lithium dendrite CNN, and outputting the safety percentage numerical value and the safety grade of the lithium battery to be detected.
The loss function and accuracy of the percentage quantization model of the lithium dendrite CNN in the training process of the percentage quantization data set of the lithium dendrite CNN of the embodiment are reduced from round to round on the training set, and finally converge, wherein the 19 th round is reduced to 0.020, and the 19 th round is reduced to 0.6221 on the testing set. The experimental results show that under the training of the model with the selected parameters, the model achieves the accuracy of 98% to 100% on the training set and the accuracy of about 98.83% on the testing set after 19 rounds of iteration. The results prove that the detection method has higher classification accuracy and lithium battery percentage quantification results. The test results and line test results in training are shown in table 1:
TABLE 1 Battery safety degree corresponding table
The lithium battery safety degree evaluation device based on lithium dendrite morphology image recognition of the embodiment comprises:
the estimation module is used for estimating the safety degree of the current state of the battery according to the lithium battery safety degree estimation method based on the lithium dendrite morphology image recognition in the embodiment;
the interval matching module is used for establishing a safety degree comparison table, wherein the safety degree comparison table is composed of a plurality of safety intervals, and the safety intervals correspond to the battery safety conditions at the current moment; matching the safety degree value obtained by the estimation module with the safety interval to obtain the battery safety condition at the current moment;
and the display module is used for displaying the safety degree value.
The estimation module and the interval matching module can be integrated in an electronic device, and specifically comprise a processor and a memory, wherein the memory stores a battery safety degree estimation method and an interval matching instruction in the embodiment, and the processor is used for calling the instruction to execute the battery safety degree estimation method and the interval matching instruction in the embodiment of the invention; the estimation module and the interval matching module may be two electronic devices, each of the two electronic devices includes a processor and a memory, a battery safety degree estimation method instruction in an embodiment is stored in the memory of the electronic device of the estimation module, the processor is configured to call the instruction to execute the battery safety degree estimation method instruction in the embodiment of the present invention, a safety degree interval matching instruction in the embodiment is stored in the memory of the electronic device of the interval matching module, and the processor is configured to call the instruction to execute the safety degree interval matching instruction in the embodiment of the present invention.
The instructions in the memory may be implemented in the form of a software functional unit and may be stored in a computer-readable storage medium when being sold or used as a standalone product, that is, a technical solution of the present invention is essentially or partially contributed to by the prior art, or a part of the technical solution may be embodied in the form of a software product, where the computer software product is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: u disk, removable hard disk, read only memory, random access memory, magnetic or optical disk, etc. for storing program codes.
In practical application, the processor can be an MSP430 singlechip, a 51 singlechip, a DSP, a TMS singlechip, an STM32 singlechip, a PIC singlechip, an AVR singlechip, an STC singlechip, a Freescale series singlechip and the like, and the singlechip can be connected with a charging and discharging source in a serial port or bus mode.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (8)
1. A lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition is characterized by comprising the following steps:
obtaining a lithium dendrite morphology image of a lithium battery to be evaluated, and preprocessing the image;
establishing a percentage quantification model of the lithium dendrite CNN, and sending the preprocessed lithium dendrite morphology image into the percentage quantification model of the lithium dendrite CNN for training;
accurately quantizing the safety percentage of the lithium battery by using the trained percentage quantization model of the lithium dendrite CNN;
establishing a safety degree comparison table, wherein the safety degree comparison table is composed of a plurality of safety intervals, and the safety intervals correspond to the battery safety conditions at the current moment; matching the obtained percentage quantitative value of the safety degree of the lithium battery with the safety interval to obtain the safety condition of the battery at the current moment;
and calculating and evaluating the percentage quantitative value and the safety grade of the safety degree of the lithium battery to be detected by utilizing the trained percentage quantitative model of the CNN of the lithium dendrite, and outputting the percentage quantitative value and the safety grade of the safety degree of the lithium battery to be detected.
2. The lithium battery safety assessment method based on lithium dendrite morphology image recognition according to claim 1, wherein the preprocessing process comprises:
obtaining a lithium dendrite morphology image of a lithium battery, and classifying the lithium dendrite morphology image;
data balance is carried out on the lithium dendrite image of the lithium battery, and the problem that the obtained lithium dendrite image has data unbalance is solved;
and carrying out numerical conversion on the lithium dendrite topography obtained through data balance to obtain a percentage quantization data set of the lithium dendrite CNN.
3. The lithium battery safety evaluation method based on lithium dendrite morphology image recognition according to claim 2, wherein the classification of the lithium dendrite morphology image is: a no lithium ion deposition topography image, a no bifurcation topography image, a cluster-like topography image, a visible distinct bifurcation structure topography image, and a visible lithium dendrite has pierced the separator topography image.
4. The lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition according to claim 1, wherein the percentage quantification model of the lithium dendrites CNN comprises 16 layers, and the specific steps from input to output are as follows in sequence: the first time is an input layer, the second layer is subjected to twice convolution of 64 convolution kernels, a pooling layer is performed for the first time, the second time is subjected to 128 convolution kernel convolution layers, the first pooling layer is performed, 512 convolution kernel convolutions are repeated twice, and the second layer sequentially passes through the pooling layer and a three-time full-connection layer.
5. The lithium battery safety evaluation method based on lithium dendrite morphology image recognition according to claim 1, wherein the lithium battery safety percentage quantification value is obtained by the following formula:
wherein, w j The weight coefficient which influences the battery safety and corresponds to the extracted jth characteristic value is adopted, n is the total number of the characteristics, f j Is the characteristic value of the j-th lithium dendrite characteristic extracted.
6. The lithium battery safety assessment method based on lithium dendrite morphology image recognition according to claim 5, wherein the characteristic value of the jth lithium dendrite characteristic is as follows:
wherein I i Is the characteristic value of the ith lithium dendrite characteristic in the image.
7. The lithium battery safety assessment method based on lithium dendrite morphology image recognition as claimed in claim 1, wherein the lithium battery safety percentage quantification value is divided into five intervals, and the five lithium battery safety percentage quantification value intervals are set to five safety levels of extreme safety, good safety, general safety, potential danger and serious danger corresponding to the safety level.
8. A lithium battery safety degree evaluation device based on lithium dendrite morphology image recognition is used for realizing the lithium battery safety degree evaluation method based on lithium dendrite morphology image recognition according to claim 1, and the method is characterized by comprising the following steps:
the estimation module is used for estimating a safety degree percentage quantification value of the current state of the battery according to the trained percentage quantification model of the lithium dendrite CNN;
the interval matching module is used for establishing a safety degree comparison table, the safety degree comparison table is composed of a plurality of safety intervals, and the safety intervals correspond to the battery safety conditions at the current moment; matching the safety degree percentage quantitative value obtained by the estimation module with the safety interval to obtain the battery safety condition at the current moment;
and the display module is used for displaying the safety degree percentage quantification value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010857310.0A CN111967190B (en) | 2020-08-24 | 2020-08-24 | Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010857310.0A CN111967190B (en) | 2020-08-24 | 2020-08-24 | Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111967190A CN111967190A (en) | 2020-11-20 |
| CN111967190B true CN111967190B (en) | 2023-01-31 |
Family
ID=73390406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010857310.0A Active CN111967190B (en) | 2020-08-24 | 2020-08-24 | Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111967190B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112750975B (en) * | 2020-12-29 | 2022-07-12 | 珠海冠宇电池股份有限公司 | Lithium battery positive plate, winding type battery cell and lithium ion battery |
| CN112767378A (en) * | 2021-01-28 | 2021-05-07 | 佛山科学技术学院 | Dense-Unet-based vitreous opacity degree rating method |
| CN113066060B (en) * | 2021-03-23 | 2023-12-19 | 湖南珞佳智能科技有限公司 | Laser directional energy precipitation image identification method of cavity convolutional neural network |
| CN113839105B (en) * | 2021-09-13 | 2024-02-06 | 中国科学院广州能源研究所 | Method for actively protecting internal short circuit of battery |
| CN114092322A (en) * | 2022-01-19 | 2022-02-25 | 清华大学 | Method and device for generating anomaly detection model and method and device for detecting anomaly |
| DE202022100786U1 (en) | 2022-02-11 | 2022-03-03 | Gaurav Kumar Arora | Intelligent processed image recognition system with artificial intelligence and machine learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102709592A (en) * | 2012-06-01 | 2012-10-03 | 中国东方电气集团有限公司 | Lithium ion secondary battery and preparation method thereof |
| CN103531839A (en) * | 2012-07-04 | 2014-01-22 | 中国科学院物理研究所 | Rechargeable metal lithium secondary battery capable of preventing from generating lithium dendrites |
| JP2016085062A (en) * | 2014-10-23 | 2016-05-19 | エンネット株式会社 | Device and method for determining battery deterioration |
| CN108107372A (en) * | 2017-12-14 | 2018-06-01 | 株洲广锐电气科技有限公司 | Accumulator health status quantization method and system based on the estimation of SOC subregions |
| CN109143083A (en) * | 2018-11-07 | 2019-01-04 | 重庆大学 | A kind of electric vehicle lithium ion battery analysis lithium diagnostic method of data-driven |
| CN109870655A (en) * | 2019-03-26 | 2019-06-11 | 上海工程技术大学 | A kind of evaluation method for lithium battery SOC |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IL239852A (en) * | 2015-07-08 | 2016-12-29 | Algolion Ltd | Lithium-ion battery safety monitoring |
| JP6567583B2 (en) * | 2017-03-15 | 2019-08-28 | 株式会社東芝 | Battery safety evaluation device, battery control device, battery safety evaluation method, program, control circuit, and power storage system |
| JP6744886B2 (en) * | 2018-03-20 | 2020-08-19 | 株式会社東芝 | Battery safety evaluation device, battery safety evaluation method, program, control circuit, and power storage system |
-
2020
- 2020-08-24 CN CN202010857310.0A patent/CN111967190B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102709592A (en) * | 2012-06-01 | 2012-10-03 | 中国东方电气集团有限公司 | Lithium ion secondary battery and preparation method thereof |
| CN103531839A (en) * | 2012-07-04 | 2014-01-22 | 中国科学院物理研究所 | Rechargeable metal lithium secondary battery capable of preventing from generating lithium dendrites |
| JP2016085062A (en) * | 2014-10-23 | 2016-05-19 | エンネット株式会社 | Device and method for determining battery deterioration |
| CN108107372A (en) * | 2017-12-14 | 2018-06-01 | 株洲广锐电气科技有限公司 | Accumulator health status quantization method and system based on the estimation of SOC subregions |
| CN109143083A (en) * | 2018-11-07 | 2019-01-04 | 重庆大学 | A kind of electric vehicle lithium ion battery analysis lithium diagnostic method of data-driven |
| CN109870655A (en) * | 2019-03-26 | 2019-06-11 | 上海工程技术大学 | A kind of evaluation method for lithium battery SOC |
Non-Patent Citations (4)
| Title |
|---|
| "Exploding E-Cigarettes: A Battery Safety Issue";SAURABH SAXENA 等;《IEEE Access》;20180330;第6卷;第21442-21466页 * |
| "计及循环寿命的锂离子电池优化使用研究";李旭玲 等;《 重庆理工大学学报(自然科学)》;20200215;第34卷(第2期);第40-46页 * |
| "锂动力电池健康度评价与估算方法的研究";李然;《中国优秀博硕士学位论文全文数据库(博士)工程科技Ⅱ辑》;20170115(第(2017)01期);第C035-7页 * |
| "锂离子动力电池热失控与安全管理研究综述";朱晓庆 等;《机械工程学报》;20200720;第56卷(第14期);第91-118页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111967190A (en) | 2020-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111967190B (en) | Lithium battery safety degree evaluation method and device based on lithium dendrite morphology image recognition | |
| CN111146514B (en) | Lithium ion battery module operation safety evaluation prediction method and system and electronic equipment | |
| CN113484787B (en) | Electrochemical impedance spectrum-based rapid sorting and recombining method for retired lithium ion battery | |
| CN113109715B (en) | Battery health condition prediction method based on feature selection and support vector regression | |
| CN111832221B (en) | Lithium battery life prediction method based on feature screening | |
| CN111257753B (en) | Battery system fault diagnosis method | |
| CN113805064A (en) | Lithium ion battery pack health state prediction method based on deep learning | |
| Takyi-Aninakwa et al. | An optimized long short-term memory-weighted fading extended Kalman filtering model with wide temperature adaptation for the state of charge estimation of lithium-ion batteries | |
| CN114818831B (en) | Bidirectional lithium ion battery fault detection method and system based on multi-source perception | |
| CN111983474A (en) | Lithium ion battery life prediction method and system based on capacity decline model | |
| CN112327192A (en) | Battery capacity diving phenomenon identification method based on curve form | |
| CN112686380A (en) | Neural network-based echelon power cell consistency evaluation method and system | |
| CN114280479A (en) | Electrochemical impedance spectrum-based rapid sorting method for retired batteries | |
| CN113447828A (en) | Lithium battery temperature estimation method and system based on Bayesian neural network | |
| CN114580496A (en) | Self-detection method for single fault of lithium ion battery | |
| CN114594378A (en) | Lithium battery SOH model training method, prediction method, device and equipment | |
| Cao et al. | Non-invasive characteristic curve analysis of lithium-ion batteries enabling degradation analysis and data-driven model construction: a review | |
| CN111983464A (en) | Lithium battery lithium separation testing method based on pure electric vehicle | |
| CN114460481A (en) | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism | |
| CN114636933B (en) | Composite neural network-based retired power battery capacity detection method and system | |
| CN113721149A (en) | Lithium battery capacity prediction method based on semi-supervised transfer learning | |
| Li et al. | State of charge estimation of lithium-ion batteries using improved BP neural network and filtering techniques | |
| Sun et al. | Internal short circuit fault diagnosis for the lithium-ion batteries with unknown parameters based on transfer learning optimized residual network by multi-label data processing | |
| Li et al. | Hybrid data-driven approach for predicting the remaining useful life of lithium-ion batteries | |
| CN117148165B (en) | Testing and analyzing method and system for polymer lithium ion battery |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |