A rapid screening method for lithium-ion battery consistency based on ultrasonic technology

Abstract

The present invention provides a method for rapid screening of lithium-ion battery consistency based on ultrasonic technology, which belongs to the technical field of battery management and cascade utilization. The method first uses ultrasonic technology to test the acoustic parameters of lithium-ion batteries, including longitudinal wave sound velocity and ultrasonic flight time; obtains the battery's electrical characteristic parameter charge and discharge capacity through charge and discharge cycle testing; establishes the relationship between the battery's acoustic characteristic parameters and electrical characteristic parameters based on a machine learning model, uses ultrasonic technology to achieve accurate estimation of the health status of lithium-ion batteries, and then achieves rapid screening of battery consistency based on cluster analysis. The present invention adopts the above-mentioned method for rapid screening of lithium-ion battery consistency based on ultrasonic technology to solve the problems of long cycle and low efficiency of existing lithium-ion battery screening methods.

Description

Technical Field

[0001] The invention relates to the technical field of battery management and recycling, and in particular to a lithium ion battery consistency rapid screening method based on ultrasonic technology. Background Art

[0002] With the continuous development of electric vehicles, energy storage systems, portable electronic devices and many other fields, lithium-ion batteries, as a high-performance, high-energy-density energy storage device, are increasingly widely used, and market demand is growing accordingly. During the life cycle of lithium-ion batteries, as the use time and frequency increase, the battery performance will gradually decline, such as capacity decay, increased internal resistance and reduced safety. However, a large number of studies have shown that even after performance degradation, many retired lithium-ion batteries still retain considerable remaining capacity and use value.

[0003] By scientifically screening and re-evaluating these retired batteries, new application scenarios can be found for them, such as energy storage systems, backup power supplies, low-speed electric vehicles, etc., so as to achieve the recycling of battery resources, extend the full life cycle of batteries, and improve resource utilization efficiency. This cascade utilization method can not only bring significant economic benefits, but also help reduce dependence on primary mineral resources, reduce environmental pollution, and promote green and low-carbon development.

[0004] In the traditional lithium-ion battery consistency screening method, the electrical characteristic parameters of the test battery are mainly used for screening and evaluation. These parameters include real-time voltage, capacity, internal resistance, etc. Although these methods can provide basic performance information of the battery, there are many limitations and shortcomings in practical applications. On the one hand, these test methods often require tedious battery charge and discharge cycles to obtain accurate electrical characteristic parameters, which not only consumes a lot of time and energy, but also causes additional losses to the battery. On the other hand, due to the complexity and non-uniformity of the internal structure of the battery, traditional testing methods are often difficult to fully and accurately reflect the internal state and health of the battery, especially the understanding of the internal microstructure and chemical reaction process of the battery is relatively limited.

[0005] In addition, traditional screening methods also face challenges in efficiency and accuracy when processing large numbers of batteries. With the large-scale development of electric vehicles and energy storage systems, the number of batteries that need to be screened has increased dramatically. Traditional screening methods often require a lot of time and resources to process these batteries, making it difficult to meet the needs of practical applications. Summary of the invention

[0006] The purpose of the present invention is to provide a lithium ion battery consistency rapid screening method based on ultrasonic technology, which solves the problems of long cycle and low efficiency of the existing lithium ion battery screening method.

[0007] To achieve the above object, the present invention provides a method for rapid screening of lithium ion battery consistency based on ultrasound technology, comprising the following steps:

[0008] Step S1: Use the ultrasonic detection system dual crystal probe to scan the lithium-ion battery on both sides to obtain acoustic characteristic parameters, including the longitudinal wave velocity. and ultrasound time of flight ;

[0009] Step S2: Perform a charge-discharge cycle test on the lithium-ion battery to obtain the charge-discharge capacity, and calculate the health status of the lithium-ion battery based on the actual discharge capacity and the rated capacity. ;

[0010] Step S3: longitudinal wave velocity and ultrasound time of flight Perform normalization processing to obtain normalized acoustic feature parameters and construct a data set;

[0011] Step S4: define the normalized acoustic feature parameters as a data matrix , will actually test the health status of the lithium-ion battery Defined as a data matrix ;

[0012] Step S5: Divide the data set into a training set and a test set. The test set accounts for 20% of the data set. Based on the gradient boosting regression algorithm, the data matrix As input, the data matrix For output, an acoustic evaluation model is established;

[0013] Step S6: using the test set data to verify the constructed acoustic evaluation model;

[0014] Step S7: Estimate the health status of the lithium-ion battery based on the established acoustic evaluation model , density cluster analysis is performed on the estimation results;

[0015] Step S8: After clustering and grouping, calculate the number of lithium-ion batteries in each group and the health status of the lithium-ion batteries in the group The standard deviation of is used as the consistency evaluation index;

[0016] Step S9, retaining the group with a standard deviation less than 0.01, and completing the rapid screening of lithium-ion battery consistency based on ultrasonic technology.

[0017] Preferably, the longitudinal wave velocity and ultrasound time of flight At 50% SOC Obtained below.

[0018] Preferably, the longitudinal wave velocity The calculation formula is as follows:

[0019] (1);

[0020] in, represents the elastic modulus; Indicates density; represents Poisson's ratio.

[0021] Preferably, ultrasound flight time The calculation formula is as follows:

[0022] (2);

[0023] in, Indicates the detection of lithium-ion battery thickness.

[0024] Preferably, the data matrix , , represents the total number of lithium-ion battery samples, Contains Longitudinal wave velocity and ultrasonic flight time of lithium-ion battery samples; data matrix , Indicates Tests on lithium-ion battery samples .

[0025] Preferably, in step S5, establishing an acoustic evaluation model comprises the following steps:

[0026] Step S51, initializing the model;

[0027] (3);

[0028] in, Represents the initial model, i.e. the target variable battery health status ;

[0029] Step S52: For each lithium-ion battery sample, calculate the residual ;

[0030] (4);

[0031] in, Before Tree pair Lithium-ion batteries The estimated results of Indicates the number of iterations;

[0032] Step S53: training decision tree The fitted residual ;

[0033] (5);

[0034] in, represents the loss function; It means to find a function that minimizes the following expression; Indicates that the current model Lithium-ion battery samples Estimated value; Represents a function in a decision tree model;

[0035] Step S54, determining the number of leaf nodes;

[0036] (6);

[0037] in, Indicates the number of leaf nodes of the decision tree; Indicates the depth of the decision tree;

[0038] Step S55: Add to formula (7), through the learning rate Weighting:

[0039] (7);

[0040] in, Indicates The prediction function of the iteration; Indicates The prediction function of the iteration;

[0041] Step S56: After rounds of iterations, the final model is the superposition of all trees:

[0042] (8);

[0043] (9);

[0044] in, Indicates The prediction function of the iteration; Represents the total number of decision trees used in the gradient boosting regression algorithm, that is, the total number of iterations.

[0045] Preferably, in step S6, the constructed acoustic evaluation model is verified using the test set data:

[0046] (10);

[0047] in, Represents the model for Lithium-ion batteries The estimated results of represents the mean square error.

[0048] Preferably, in step S7, based on the established acoustic evaluation model, the health status of the lithium-ion battery is estimated and obtained. , density cluster analysis is performed on the estimation results, including the following steps:

[0049] Step S71, determining the core point of the lithium-ion battery;

[0050] Step S72: For any lithium-ion battery core point, density clustering will expand the current group to other close points. When the new point is a lithium-ion battery core point, it will be used to propagate the group; when the new point is not a lithium-ion battery core point, it will be marked as a non-core point.

[0051] Step S73: All points not assigned to any group will be marked as discrete points.

[0052] Preferably, the health status of lithium-ion batteries The standard deviation of is calculated as follows:

[0053] (11);

[0054] in, Represents the estimate of all lithium-ion battery samples in the group The average of the values; Represents the number of lithium-ion battery samples in the group after screening; Represents the standard deviation of the health state of a lithium-ion battery.

[0055] Therefore, the present invention adopts the above-mentioned lithium ion battery consistency rapid screening method based on ultrasonic technology, and the beneficial technical effects are as follows:

[0056] (1) High efficiency: Ultrasonic nondestructive testing technology is used to obtain the longitudinal wave speed and ultrasonic flight time of the battery as characteristic parameters for consistency screening. Compared with the traditional screening method of obtaining electrical characteristic parameters such as capacity, voltage, and internal resistance, the test cycle is short, faster, more efficient, and convenient for practical engineering applications.

[0057] (2) Accuracy: The present invention adopts the gradient boosting regression algorithm (GBR) to establish a relationship model between battery acoustic characteristic parameters and electrical characteristic parameters, which has higher estimation accuracy.

[0058] (3) Self-adaptation: The present invention is based on ultrasonic technology and adopts a density clustering algorithm to complete the consistency screening of batteries. It can automatically identify batteries with similar characteristics without preset grouping and can adapt to the nonlinearity and diversity of battery parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a flow chart of a method for rapid screening of lithium-ion battery consistency based on ultrasonic technology of the present invention;

[0060] Figure 2 This is a schematic diagram of ultrasonic scanning of 18650 cylindrical batteries;

[0061] Figure 3 For estimation Result verification and error analysis;

[0062] Figure 4 This is the consistency screening result of 18650 lithium-ion batteries. DETAILED DESCRIPTION

[0063] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0064] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0065] Embodiment 1

[0066] like Figure 1 As shown in the figure, it is a flow chart of a lithium-ion battery consistency rapid screening method based on ultrasonic technology. In this embodiment, 60 18650 lithium-ion batteries discarded from electric bicycles are used, and they are numbered 1, 2, 3, ..., 60 respectively. SOC Under normal conditions, ultrasonic scanning is performed on the battery to obtain the longitudinal wave sound velocity and ultrasonic flight time of the battery and normalize them; charge and discharge cycle tests are performed on the battery to obtain the actual discharge capacity and calculate its .

[0067] like Figure 2 The figure shows a schematic diagram of an 18650 cylindrical battery receiving ultrasonic scanning. The piezoelectric chip is simplified into a small probe to perform a full-scale scan of the battery. The penetrating wave collects the internal state, structure, and defect information of the battery.

[0068] The longitudinal wave speed of sound and ultrasonic flight time are closely related to the battery status, and can effectively reflect the physical properties and structural changes of the internal materials of the battery. They have a nonlinear relationship with the battery's electrical characteristic parameters (such as charge and discharge capacity, battery health status, etc.). With the help of machine learning and other algorithms, the relationship between the longitudinal wave speed of sound, ultrasonic flight time and battery health status can be accurately established.

[0069] Based on the obtained battery longitudinal wave speed, ultrasonic flight time and other acoustic characteristic parameters and , the gradient boosting regression algorithm is used to establish the battery acoustic characteristic parameters and 80% of the batteries, i.e. 48 batteries, are randomly selected from the 60 batteries to construct the battery acoustic characteristic parameters and The relationship model training is as follows:

[0070] Step S51, initializing the model;

[0071] (1);

[0072] in, Represents the initial model, i.e. the target variable battery health status ;

[0073] Step S52: For each lithium-ion battery sample, calculate the residual ;

[0074] (2);

[0075] in, Before Tree pair Lithium-ion batteries The estimated results of Indicates Li-ion battery sample residuals, Indicates the number of iterations;

[0076] Step S53: training decision tree The fitted residual ;

[0077] (3);

[0078] in, Indicates The decision tree of the iteration; represents the loss function; It means to find a function that minimizes the following expression; Indicates that the current model Lithium-ion battery samples Estimated value, h It represents a function in a decision tree model for estimating the health status of lithium-ion battery samples;

[0079] Step S54, determining the number of leaf nodes;

[0080] (4);

[0081] in, Indicates the number of leaf nodes of the decision tree; Indicates the depth of the decision tree;

[0082] Step S55: Decision tree Add to formula (5), through the learning rate Weighting:

[0083] (5);

[0084] in, Indicates The prediction function of the iteration; Indicates The prediction function of the iteration;

[0085] Step S56: After rounds of iterations, the final model is the superposition of all trees:

[0086] (6);

[0087] (7);

[0088] in, Indicates The prediction function of the iteration; Represents the total number of decision trees used in the gradient boosting regression algorithm, that is, the total number of iterations;

[0089] Step 7: Model validation: Use the test set data to validate the constructed model. Calculate the mean square error (MSE) between the model estimation results and the actual values ​​to evaluate the accuracy of the model:

[0090] (8);

[0091] in, Represents the model for Lithium-ion batteries The estimated results of represents the mean square error.

[0092] Then, the remaining 20% ​​of the batteries, i.e., 12 batteries, are used to verify the above-mentioned model. Figure 3 As shown, by constructing a relationship model, the longitudinal wave speed of sound and ultrasonic flight time can be used to accurately estimate the battery , estimated The mean square error is 6.36e-5.

[0093] For the estimated Further cluster analysis involves parameter calculation process as follows:

[0094] (1) Neighborhood radius (eps):

[0095] The eps parameter defines the size of the neighborhood, that is, a battery sample is considered to be an approximate individual of another battery sample within the eps distance. In this example, it can be interpreted as the size of the neighborhood of the two batteries. The range is considered similar based on the similarity of battery performance. The calculation process is as follows:

[0096] 1) Calculate the nearest neighbor distance: For each battery sample point in the data set, calculate its distance to the kth nearest neighbor. Here, we use Euclidean distance as the metric. For two battery sample points and , which is Euclidean distance in dimension The calculation formula is:

[0097] (9);

[0098] in, Represents sample points of value; Represents sample points of value.

[0099] 2) Construct k-distance: For each battery sample point, we record its distance to the kth nearest neighbor and arrange them in ascending order.

[0100] 3) Select the eps value: In the k-distance, find the point where the distance value suddenly increases. This point usually marks the transition from a low-density area to a high-density area. The k-distance value corresponding to this transition point is the value of the neighborhood radius eps, that is:

[0101] (10);

[0102] in, Represents the value of the neighborhood radius; Indicates the sorted In this embodiment, the value of the neighborhood radius is 1.15e-3.

[0103] (2) Minimum number of samples:

[0104] Defines the minimum number of samples required to form a dense region. In this example, the minimum number of samples is 4, indicating that a battery sample core point can be considered a core point only if there are at least 4 similar batteries. In battery consistency screening, this ensures that only when a certain number of batteries show similar values, they are classified into the same group, thus improving the reliability of clustering.

[0105] (3) The process of clustering analysis of batteries based on density clustering algorithm is as follows:

[0106] 1) Identification of battery core points:

[0107] The algorithm first determines which points are battery core points. If the number of approximate battery samples of a point is greater than or equal to the minimum number of samples, the point is considered to be a battery core point. Only when the value is similar to at least 4 other batteries, it is considered a core battery and participates in the screening of the group.

[0108] 2) Group expansion:

[0109] For each battery core point, the algorithm will expand the current group to other nearby points. If the new point is a battery core point, it will be used to propagate the group; if not, it will be marked as a non-core point. Similar to the core battery, it will be classified into the same group, indicating that these batteries are consistent in performance.

[0110] 3) Identification of outliers:

[0111] All points not assigned to any group will be marked as discrete points. In this embodiment, batteries that are inconsistent with the performance of most batteries can be identified.

[0112] Finally, a group label is assigned to each battery sample, and battery sample points with the same label belong to the same group, such as Figure 4 As shown in Table 1, the consistency screening results of 60 18650 batteries in this embodiment are obtained. A total of 47 batteries are screened out and divided into 3 groups, with group 0 containing 38 batteries, group 1 containing 5 batteries, and group 2 containing 4 batteries. The standard deviation of SOH between batteries in the group is calculated, which is 3.2e-03, 4.48e-04, and 8.89e-04 respectively. The remaining 13 batteries are eliminated, which are discrete samples (group-1). The consistency grouping results of 18650 lithium-ion batteries in this embodiment are specifically shown in Table 1.

[0113] The standard deviation is calculated as follows:

[0114] (11);

[0115] in, Represents the estimate of all lithium-ion battery samples in the group The average value of Represents the number of lithium-ion battery samples in the group after screening, Represents the standard deviation of the health status of the lithium-ion batteries in the group.

[0116] Table 1 18650 lithium-ion battery consistency grouping results

[0117]

[0118] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0119] Therefore, the present invention adopts the above-mentioned lithium-ion battery consistency rapid screening method based on ultrasonic technology, which has the advantages of high efficiency and non-destructiveness, and solves the problem that the traditional screening method has a long test cycle and limited engineering application. The present invention not only improves the screening efficiency, but also provides a scientific decision-making basis for the recycling and reuse of batteries, which helps to promote the development of battery recycling economy.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A rapid screening method for lithium-ion battery consistency based on ultrasonic technology, characterized in that: The following steps are involved: Step S1: Use the ultrasonic detection system dual crystal probe to scan the lithium-ion battery on both sides to obtain acoustic characteristic parameters, including the longitudinal wave velocity. and ultrasound time of flight ; Step S2: Perform a charge-discharge cycle test on the lithium-ion battery to obtain the charge-discharge capacity, and calculate the health status of the lithium-ion battery based on the actual discharge capacity and the rated capacity. ; Step S3: longitudinal wave velocity and ultrasound time of flight Perform normalization processing to obtain normalized acoustic feature parameters and construct a data set; Step S4: define the normalized acoustic feature parameters as a data matrix , will actually test the health status of the lithium-ion battery Defined as a data matrix ; Step S5: Divide the data set into a training set and a test set. The test set accounts for 20% of the data set. Based on the gradient boosting regression algorithm, the data matrix As input, the data matrix For output, an acoustic evaluation model is established; Step S6: using the test set data to verify the constructed acoustic evaluation model; Step S7: Estimate the health status of the lithium-ion battery based on the established acoustic evaluation model , density cluster analysis is performed on the estimation results; Step S8: After clustering and grouping, calculate the number of lithium-ion batteries in each group and the health status of the lithium-ion batteries in the group The standard deviation of is used as the consistency evaluation index; Step S9, retaining the group with a standard deviation less than 0.01, and completing the rapid screening of lithium-ion battery consistency based on ultrasonic technology; Longitudinal wave speed and ultrasound time of flight At 50% SOC Obtained under Longitudinal wave speed The calculation formula is as follows: （1）； in, represents the elastic modulus; Indicates density; represents Poisson's ratio.

2. According to claim 1, a rapid screening method for lithium-ion battery consistency based on ultrasonic technology is characterized in that: Ultrasound time of flight The calculation formula is as follows: （2）； in, Indicates the detection of lithium-ion battery thickness.

3. The method for rapid screening of lithium-ion battery consistency based on ultrasonic technology according to claim 2, characterized in that: Data Matrix , , represents the total number of lithium-ion battery samples, Contains Longitudinal wave velocity and ultrasonic flight time of lithium-ion battery samples; data matrix , Indicates Lithium-ion battery samples tested .

4. The method for rapid screening of lithium-ion battery consistency based on ultrasonic technology according to claim 3, characterized in that: In step S5, an acoustic evaluation model is established, including the following steps: Step S51, initializing the model; （3）； in, Represents the initial model, i.e. the target variable battery health status ; Step S52: For each lithium-ion battery sample, calculate the residual ; （4）； in, Before Tree pair Lithium-ion batteries The estimated results of Indicates the number of iterations; Step S53: training decision tree The fitted residual ; （5）； in, represents the loss function; It means to find a function that minimizes the following expression; Indicates that the current model Lithium-ion battery samples Estimated value; Represents a function in a decision tree model; Step S54, determining the number of leaf nodes; （6）； in, Indicates the number of leaf nodes of the decision tree; Indicates the depth of the decision tree; Step S55: Add to formula (7), through the learning rate Weighting: （7）； in, Indicates The prediction function of the iteration; Indicates The prediction function of the iteration; Step S56: After rounds of iterations, the final model is the superposition of all trees: （8）； （9）； in, Indicates The prediction function of the iteration; Represents the total number of decision trees used in the gradient boosting regression algorithm, that is, the total number of iterations.

5. The method for rapid screening of lithium-ion battery consistency based on ultrasonic technology according to claim 4, characterized in that: In step S6, the constructed acoustic evaluation model is verified using the test set data: （10）； in, Represents the model for Lithium-ion batteries The estimated results of represents the mean square error.

6. The method for rapid screening of lithium-ion battery consistency based on ultrasonic technology according to claim 5, characterized in that: In step S7, the health status of the lithium-ion battery is estimated based on the established acoustic evaluation model. , density cluster analysis is performed on the estimation results, including the following steps: Step S71, determining the core point of the lithium-ion battery; Step S72: for any lithium-ion battery core point, density clustering will expand the current group to other close points. When the new point is a lithium-ion battery core point, it will be used to propagate the group; when the new point is not a lithium-ion battery core point, it will be marked as a non-core point. Step S73: All points not assigned to any group will be marked as discrete points.

7. The method for rapid screening of lithium-ion battery consistency based on ultrasonic technology according to claim 6, characterized in that: Lithium-ion battery health status The standard deviation of is calculated as follows: （11）； in, Represents the estimate of all lithium-ion battery samples in the group The average of the values; Represents the number of lithium-ion battery samples in the group after screening; Represents the standard deviation of the health state of a lithium-ion battery.

Images