Ternary lithium battery pole piece performance test method and system fused with machine learning

Abstract

The invention provides a ternary lithium battery pole piece performance testing method and system fused with machine learning, and relates to the technical field of performance detection.The method comprises the steps that multi-point thickness testing is conducted on a ternary lithium battery pole piece, multiple pole piece thicknesses are obtained, and multiple discharge slopes are obtained through testing; performing discharge slope prediction according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and performing pole piece thickness prediction according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses; randomly combining the thicknesses of the plurality of pole pieces and the thicknesses of the plurality of predicted pole pieces, calculating to obtain a thickness confidence coefficient, randomly combining the plurality of discharge slopes and the plurality of predicted discharge slopes, and calculating to obtain a discharge confidence coefficient; and according to the thickness confidence coefficient and the discharge confidence coefficient, calculating a performance test result of the ternary lithium battery pole piece in combination with the thicknesses of the pole pieces and the discharge slopes. The technical problem of low reliability of performance test data of the ternary lithium battery pole piece in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of performance detection, and in particular to a ternary lithium battery pole piece performance test method and system fusing machine learning. BACKGROUND

[0002] The thickness uniformity and discharge performance of a ternary lithium battery pole piece are core key indicators directly determining the energy density, cycle life and safety of the battery.

[0003] However, conventional ternary lithium battery pole piece performance tests mostly rely on random single-point detection mode, which is strongly affected by the randomness of the detection points and the limited amount of data, and is prone to result in low test data reliability, and thus the test results have low reference value for the true performance of the pole piece.

[0004] Therefore, there is an urgent need for a ternary lithium battery pole piece performance precision test method that can effectively avoid the limitations of random single-point detection. SUMMARY

[0005] The present application provides a ternary lithium battery pole piece performance test method and system fusing machine learning to solve the technical problem of low test data reliability of ternary lithium battery pole pieces in the prior art.

[0006] The technical solution of the present application to solve the above technical problem is as follows:

[0007] In a first aspect, the present application provides a ternary lithium battery pole piece performance test method fusing machine learning, comprising:

[0008] Performing multi-point thickness testing on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and testing to obtain a plurality of discharge slopes of the ternary lithium battery pole piece;

[0009] Respectively predicting the discharge slope according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and predicting the pole piece thickness according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses;

[0010] Randomly combining the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to calculate a thickness confidence, and randomly combining the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate a discharge confidence, wherein the confidence is calculated according to the number of tests and the prediction accuracy;

[0011] According to the thickness confidence and the discharge confidence, combining the plurality of pole piece thicknesses and the plurality of discharge slopes to calculate a performance test result of the ternary lithium battery pole piece.

[0012] In a second aspect, the present application provides a ternary lithium battery pole piece performance test system fusing machine learning, comprising:

[0013] The thickness test module is used for multi-point thickness test on the ternary lithium battery pole piece, obtains a plurality of pole piece thicknesses, and tests a plurality of discharge slopes of the ternary lithium battery pole piece;

[0014] The parameter prediction module is used for respectively performing discharge slope prediction according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and performing pole piece thickness prediction according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses;

[0015] The confidence analysis module is used for randomly combining the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to obtain thickness confidence, and randomly combining the plurality of discharge slopes and the plurality of predicted discharge slopes to obtain discharge confidence, wherein the confidence is calculated according to the test quantity and the prediction accuracy;

[0016] The fusion output module is used for calculating the performance test result of the ternary lithium battery pole piece according to the thickness confidence and the discharge confidence and in combination with the plurality of pole piece thicknesses and the plurality of discharge slopes.

[0017] The present application has the following beneficial effects:

[0018] Compared with the prior art, firstly, the present application performs multi-point thickness test on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and tests a plurality of discharge slopes of the ternary lithium battery pole piece, thereby providing a reliable basis for subsequent machine learning prediction and confidence analysis by collecting representative and stable data. Secondly, the present application respectively performs discharge slope prediction according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and performs pole piece thickness prediction according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses, thereby realizing bidirectional prediction of pole piece thickness→predicted discharge slope and discharge slope→predicted pole piece thickness through a machine learning model, verifying data consistency through bidirectional mapping, and providing a prediction benchmark for subsequent confidence calculation. Thirdly, the present application randomly combines the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to obtain thickness confidence, and randomly combines the plurality of discharge slopes and the plurality of predicted discharge slopes to obtain discharge confidence, thereby converting the consistency of measured data and predicted data into a quantifiable reliability index, fully considering the influence of test scale and model reliability on confidence, and finally outputting the thickness confidence and the discharge confidence to provide an explicit credibility label for the pole piece performance test result, thereby solving the pain point that the referenceability of traditional test cannot be quantified. Finally, the present application calculates the performance test result of the ternary lithium battery pole piece according to the thickness confidence and the discharge confidence and in combination with the plurality of pole piece thicknesses and the plurality of discharge slopes, thereby quantifying the influence of thickness, discharge performance and data reliability on the performance of the ternary lithium battery pole piece, making the performance evaluation comprehensive and reliable, and providing a directly applicable quantifiable basis for pole piece quality grading and process optimization.

[0019] By multi-point testing on the ternary lithium battery pole piece, a plurality of pole piece thicknesses and corresponding discharge slopes are obtained, bidirectional prediction of pole piece thickness to predicted discharge slope and discharge slope to predicted pole piece thickness is realized according to a machine learning model, a cross verification system of measured data and predicted data is formed, basic confidence is calculated by randomly combining measured and predicted data, and double correction is carried out in combination with test quantity (data sufficiency) and prediction accuracy (model reliability), so as to accurately quantify thickness confidence and discharge confidence, and finally, the test results of the pole piece comprehensive performance are obtained by fusing multi-dimensional indexes. In this way, the accidental error of the traditional single-point detection is effectively avoided through multi-measurement point data acquisition and bidirectional prediction, the data representativeness and relevance are improved, the traceability evaluation of the reliability of the test results is realized through confidence quantification, the representativeness and reliability of the performance test data of the ternary lithium battery pole piece are improved, and high-precision, quantifiable and reliable technical support is provided for quality control, production process optimization and battery performance improvement of the ternary lithium battery pole piece. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a ternary lithium battery pole piece performance test method fusing machine learning is provided.

[0021] Figure 2 A structure diagram of a ternary lithium battery pole piece performance test system fusing machine learning is provided.

[0022] In the drawings, the components represented by the numbers are as follows:

[0023] The thickness test module 11, the parameter prediction module 12, the confidence analysis module 13 and the fusion output module 14. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0025] In the description of the present application, the terms "first" and "second" are used only for description purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0026] In the description of the present application, the term "for example" is used to mean "serving as an example, instance, or illustration." Any embodiment described as "for example" in the present application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the application. In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present application. It will be apparent, however, to one skilled in the art that the present application can be practiced without using these specific details. In other instances, well-known structures and processes are not elaborated upon in order to avoid unnecessary detail, which can obscure the description of the present application. Thus, the present application is not intended to be limited by the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0027] As shown in Embodiment I, Figure 1 The present application provides a method for testing the performance of a ternary lithium battery pole piece by fusing machine learning, which comprises:

[0028] S10: Multi-point thickness testing is performed on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and a plurality of discharge slopes of the ternary lithium battery pole piece are tested and obtained.

[0029] Traditional performance testing of ternary lithium battery pole pieces relies on random single-point thickness measurement, which is affected by the randomness of the measurement points and the insufficient amount of data samples, and is prone to problems such as insufficient representativeness and large fluctuations of the data, ultimately resulting in low reliability and reference value of the test results.

[0030] To solve the above problems, the present application performs multi-point thickness testing on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and a plurality of discharge slopes of the ternary lithium battery pole piece are tested and obtained.

[0031] Specifically, step S10 in the method comprises:

[0032] Obtaining a test number;

[0033] Randomly selecting test positions on the ternary lithium battery pole piece, performing multi-point thickness testing, and obtaining a plurality of pole piece thicknesses;

[0034] Performing constant-current discharge testing on the ternary lithium battery to obtain a discharge curve at a middle SOC;

[0035] According to the test number, the discharge curve is divided, and the slopes of the divided plurality of discharge curve segments are obtained as a plurality of discharge slopes.

[0036] In the embodiments of the present application, the test quantity is first acquired. Specifically, the test quantity refers to the number of positions randomly selected on the ternary lithium battery pole piece for multi-point thickness testing, i.e., the number of sampling points for pole piece thickness testing, such as 5, 10, etc. The test quantity directly affects the representativeness of the data. Too few test quantities are easily affected by single-point accidental errors (such as thickness abnormalities caused by local defects of the pole piece). Too many test quantities will increase the testing cost and time. Therefore, a reasonable test quantity can be preset according to factors such as the size of the actual ternary lithium battery pole piece, the stability of the production process, etc., to ensure that the test quantity can reflect the overall characteristics and also take into account the testing efficiency.

[0037] Secondly, the test quantity of test positions is randomly selected on the ternary lithium battery pole piece, and multi-point thickness testing is performed to obtain a plurality of pole piece thicknesses. The thickness testing can be performed by a laser thickness gauge, a micrometer, etc. For example, according to the test quantity (such as 5), 5 different positions are randomly selected on the ternary lithium battery pole piece, and the thickness of the 5 different positions is measured by a laser thickness gauge to obtain 5 pole piece thicknesses, for example, 122 μm, 119 μm, 124 μm, 118 μm, and 123 μm. In this way, by randomly selecting the test quantity of test positions, subjective bias caused by human selection can be avoided, the sampling can cover different areas of the pole piece, the overall distribution characteristics of the thickness of the ternary lithium battery pole piece can be truly reflected, and finally a plurality of pole piece thicknesses can be obtained, which can effectively reduce the accidental error of single-point testing.

[0038] Again, the ternary lithium battery is subjected to constant current discharge test to obtain the discharge curve of the middle SOC. Specifically, the constant current discharge test refers to fixing the discharge current and testing the change of the voltage with the capacity, which is to exclude the interference of current fluctuation on the change of the voltage, and to ensure the stability and comparability of the discharge curve; the SOC (State of Charge, state of charge) is a core parameter for measuring the remaining capacity of the battery, usually expressed in percentage, 0% is completely discharged, and 100% is completely charged, and its physical nature is the ratio of the current remaining capacity to the rated capacity, which directly reflects the available energy state of the battery; the discharge curve of the middle SOC refers to the discharge curve of the state of charge of 20% to 80%, and the discharge curve of the middle SOC is obtained because the voltage changes with the capacity at high SOC (such as >80%) or low SOC (such as <20%) often presents nonlinear characteristics, such as a flat voltage platform at high SOC, a rapid voltage drop at low SOC, an unstable slope and a large influence of the environment (such as temperature), while in the middle SOC range, the relationship between the battery voltage and the remaining capacity presents a high linearity, a good repeatability of the slope, and can more stably reflect the electrochemical performance of the battery and reduce the influence of accidental interference. Exemplarily, when the ternary lithium battery is subjected to constant current discharge test, the discharge curve is drawn with the discharge capacity as the horizontal axis and the voltage as the vertical axis, the curve segment corresponding to the SOC interval (20% to 80%) in the discharge curve is extracted as the discharge curve of the middle SOC.

[0039] Finally, the discharge curve is divided according to the number of tests, and the slopes of the divided multiple discharge curve segments are obtained as multiple discharge slopes. Specifically, the discharge curve of the middle SOC is evenly divided into n continuous segments according to the number of tests (n), and the change rate of the voltage to the capacity (i.e. the slope) of each segment is calculated to obtain n discharge slopes. The discharge curve is divided because although the discharge curve in the middle SOC interval is stable as a whole, the slopes of different subintervals may still have slight differences, and multiple discharge slopes can comprehensively reflect the discharge performance changes of the battery in the middle SOC interval, avoid the inability of a single slope to capture local characteristics, and further reduce the accidentalness of the data.

[0040] In summary, compared with the prior art, the present application tests the multiple thicknesses of the ternary lithium battery pole piece, and obtains multiple discharge slopes of the ternary lithium battery pole piece. In this way, by collecting representative and stable data, a reliable foundation is provided for subsequent machine learning prediction and confidence analysis.

[0041] S20: respectively according to the multiple thicknesses of the pole piece, the discharge slope prediction is carried out to obtain multiple predicted discharge slopes, and the thickness of the pole piece is predicted according to the multiple discharge slopes to obtain multiple predicted thicknesses of the pole piece;

[0042] In the performance test of ternary lithium battery pole piece, there is a certain correlation between the pole piece thickness and the discharge slope. Specifically, the pole piece is the core reaction area of lithium ion intercalation / deintercalation, and the pole piece thickness directly determines the diffusion distance of lithium ions in the electrode material. A thicker pole piece will lengthen the ion migration distance, causing the rate of voltage decline with capacity to increase during discharge, resulting in an increase in discharge slope. A thinner pole piece has a shorter diffusion path and smaller resistance, resulting in a more gradual discharge slope. At the same time, a pole piece that is too thick is prone to insufficient wetting or uneven conduction, which reduces the effective reaction area and further exacerbates the fluctuation of the discharge slope. A moderate thickness can improve the utilization rate of active materials and ensure the stability of the discharge curve. Therefore, based on this clear physical correlation, the discharge slope can be predicted by the pole piece thickness, and the pole piece thickness can be inferred from the discharge slope, and the abnormal data can be accurately identified and analyzed through bidirectional prediction.

[0043] To solve the above problems, the present application predicts the discharge slope according to a plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and predicts the pole piece thickness according to a plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses.

[0044] Specifically, step S20 in the method comprises:

[0045] calling a discharge slope predictor;

[0046] inputting the plurality of pole piece thicknesses into the discharge slope predictor respectively to obtain a plurality of predicted discharge slopes through prediction output;

[0047] calling a pole piece thickness predictor;

[0048] inputting the plurality of discharge slopes into the pole piece thickness predictor respectively to obtain a plurality of predicted pole piece thicknesses through prediction output.

[0049] In the embodiment of the present application, a discharge slope predictor based on machine learning is first called. The discharge slope predictor can predict the theoretical discharge slope based on the measured data of the pole piece thickness.

[0050] Secondly, a plurality of pole piece thicknesses obtained through actual testing are input into the discharge slope predictor respectively to obtain a plurality of predicted discharge slopes through prediction output. This is because the pole piece thickness directly affects the ion diffusion path length and the utilization rate of active materials. The discharge slope predictor can output the predicted discharge slope that conforms to the theoretical law by learning the correlation between the pole piece thickness and the discharge slope in the historical data.

[0051] Again, the pole piece thickness predictor is called, wherein the pole piece thickness predictor is constructed based on machine learning, and the training process thereof is the same as that of the discharge slope predictor. The pole piece thickness predictor is based on the measured data of the discharge slope to back-calculate the theoretical pole piece thickness, and forms a double-verification closed loop with the discharge slope predictor. Through the cross-prediction of the two models, the influence of single-model error can be reduced.

[0052] Finally, the plurality of discharge slopes obtained through actual testing are respectively input into the pole piece thickness predictor, and a plurality of predicted pole piece thicknesses are obtained through prediction output. The pole piece thickness predictor can back-calculate the theoretical thickness that leads to the performance by learning the correlation between the discharge slope and the pole piece thickness in the historical data.

[0053] In this way, the measured pole piece thickness is input into the discharge slope predictor, and the theoretical predicted discharge slope is output through prediction. The measured discharge slope is input into the pole piece thickness predictor, and the theoretical predicted pole piece thickness is output through prediction. The consistency of the measured pole piece thickness and the predicted pole piece thickness, and the measured discharge slope and the predicted discharge slope can be compared. If the deviation is too large, it is prompted that the measured data may have accidental errors, which provides an abnormal identification basis for subsequent confidence calculation.

[0054] Further, the “calling the discharge slope predictor” comprises:

[0055] According to the historical test data of the ternary lithium battery pole piece, a sample pole piece thickness set and a constant-current discharge curve slope of the ternary lithium battery under different sample pole piece thicknesses are collected, and a sample discharge slope set is labeled and obtained;

[0056] A discharge slope predictor based on machine learning is constructed;

[0057] The sample pole piece thickness set and the sample discharge slope set are used for supervised training of the discharge slope predictor, and the training is completed after the test converges;

[0058] The converged discharge slope predictor is configured in a cloud server.

[0059] In the embodiments of the present application, first, according to the historical test data of the ternary lithium battery pole piece, a sample pole piece thickness set and a constant-current discharge curve slope of the ternary lithium battery under different sample pole piece thicknesses are collected, and a sample discharge slope set is labeled and obtained, forming a one-to-one corresponding pole piece thickness-constant-current discharge curve slope sample pair. For example, if a certain sample pole piece thickness is 120 μm, the slope on the SOC discharge curve thereof is-0.005 V / mAh, and it is labeled as (120 μm, -0.005 V / mAh).

[0060] Secondly, a discharge slope predictor based on machine learning is constructed. Exemplarily, since the pole piece thickness has a strong linear relationship with the discharge slope, a neural network can be used to construct the discharge slope predictor, mainly composed of an input layer, a hidden layer, and an output layer architecture: the input layer receives the pole piece thickness data, the first hidden layer has 32 neurons for preliminary nonlinear transformation of the input features, the second hidden layer has 64 neurons with 20% Dropout to strengthen nonlinear fitting and suppress overfitting, and the third hidden layer has 32 neurons for compression and integration of high-order features to provide a precise mapping basis for the output layer. Finally, a single neuron in the output layer is linearly activated to output the predicted discharge slope value.

[0061] Thirdly, the discharge slope predictor is supervised trained by using the sample pole piece thickness set and the sample discharge slope set, and the training is completed after the test converges. Exemplarily, the training process of the discharge slope predictor can be realized through the following technical path: 1. Data preparation: divide the sample pole piece thickness set and the sample discharge slope set into a training set, a validation set, and a test set according to 7:1.5:1.5, the training set is used for model parameter learning, the validation set is used for generalization ability evaluation, the test set is used for testing the accuracy of the model, and the sample pole piece thickness data is standardized by Z-score to eliminate the dimension difference. 2. Model training: use the sample pole piece thickness in the training set as the input feature, and use the corresponding sample discharge slope as the supervised label. Use mean square error (MSE) as the loss function to quantify the prediction deviation. Through the Adam optimizer, the model weight parameters are iteratively optimized. At the same time, early stopping strategy (when the validation set loss does not decrease for 10 consecutive rounds, terminate the training) and L2 regularization are introduced to prevent overfitting. During the training process, the performance of the validation set is monitored in real time. When the number of iterations reaches the pre-set maximum threshold (such as 500 rounds) or the prediction accuracy on the validation set is ≥95%, it is determined that the model converges, and the trained discharge slope predictor is obtained.

[0062] Finally, the converged discharge slope predictor is configured in the cloud server, so that the discharge slope predictor can be remotely called through the network interface, without the need for repeated training on the local device, reducing the hardware resource demand. The cloud server can centrally manage the discharge slope predictor, and when new historical data is accumulated or prediction deviation is found, the model can be retrained and updated on the cloud to ensure that the model is always optimized based on the latest data and maintains high accuracy.

[0063] In summary, compared with the prior art, the present application respectively predicts the discharge slope according to multiple pole piece thicknesses, obtains multiple predicted discharge slopes, and predicts the pole piece thickness according to multiple discharge slopes to obtain multiple predicted pole piece thicknesses. In this way, the machine learning model realizes the cross-prediction of pole piece thickness→discharge slope and discharge slope→pole piece thickness, and verifies the data consistency through bidirectional mapping, providing a prediction benchmark for subsequent confidence calculation.

[0064] S30: randomly combine the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to obtain a plurality of pole piece thickness groups, respectively calculate the similarity and calculate the mean value to obtain a basic thickness confidence, randomly combine the plurality of discharge slopes and the plurality of predicted discharge slopes to obtain a plurality of discharge slope groups, respectively calculate the similarity and calculate the mean value to obtain a basic slope confidence, wherein the confidence is calculated according to the test number and the prediction accuracy;

[0065] The conventional method cannot quantify the reliability of the test data, resulting in a lack of objective evaluation criteria for the credibility of the test results, and misjudgments may occur, such as data that is consistent on the surface but actually lacks representativeness, or even biased conclusions based on accidental error data. Therefore, a confidence quantification system needs to be constructed to accurately quantify the reliability of the test data and solve the problem of the lack of reliability evaluation in the conventional method.

[0066] To solve the above problems, the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses are randomly combined to obtain a plurality of pole piece thickness groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic thickness confidence, the plurality of discharge slopes and the plurality of predicted discharge slopes are randomly combined to obtain a plurality of discharge slope groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic slope confidence, and the confidence is calculated according to the test number and the prediction accuracy.

[0067] Specifically, step S30 in the method includes:

[0068] The plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses are randomly combined to obtain a plurality of pole piece thickness groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic thickness confidence;

[0069] The plurality of discharge slopes and the plurality of predicted discharge slopes are randomly combined to obtain a plurality of discharge slope groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic slope confidence;

[0070] The thickness confidence and the discharge confidence are calculated according to the test number and the prediction accuracy.

[0071] In the embodiment of the application, the plurality of pole piece thicknesses obtained by actual measurement and the plurality of predicted pole piece thicknesses predicted by the pole piece thickness predictor are randomly combined to obtain a plurality of pole piece thickness groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic thickness confidence, and the plurality of discharge slopes and the plurality of predicted discharge slopes are randomly combined to obtain a plurality of discharge slope groups, the similarity is respectively calculated and the mean value is calculated to obtain a basic slope confidence, wherein the influence of single-point error can be diluted by random combination, and the consistency of the overall data can be more objectively reflected. For example, the similarity calculation can use the normalized value of cosine similarity or Euclidean distance, for example:

[0072] ;

[0073] wherein the similarity value ranges from 0 to 1, 1 indicates that the actual thickness of the electrode plate and the predicted thickness of the electrode plate are completely consistent, and 0 indicates that the actual thickness of the electrode plate and the predicted thickness of the electrode plate are completely irrelevant, for example, if the actual thickness of the electrode plate of the ternary lithium battery is 120 μm, and the predicted thickness of the electrode plate predicted by the electrode thickness predictor is 118 μm, then the similarity of this electrode thickness group is = 1 - (|120-118|) / 120 = 0.983, and thus, the similarity of a plurality of electrode thickness groups is calculated in the same way, and the arithmetic mean is calculated, for example, 0.98, to obtain the basic thickness confidence, which can reflect the overall consistency between the actual thickness of the electrode plate and the predicted thickness of the electrode plate predicted by the electrode thickness predictor, and the greater the basic thickness confidence, the stronger the overall consistency between the actual thickness of the electrode plate and the predicted thickness of the electrode plate, that is, the more reliable the data.

[0074] Secondly, a plurality of discharge slopes obtained by actual measurement and a plurality of predicted discharge slopes predicted by the discharge slope predictor are randomly combined to obtain a plurality of discharge slope groups, the similarity is calculated respectively, and the mean value is calculated to obtain the basic slope confidence. Exemplarily, the similarity is calculated according to the same logic and method as the basic thickness confidence, for example:

[0075] ;

[0076] wherein the similarity value ranges from 0 to 1, 1 indicates that the actual thickness of the electrode plate and the predicted thickness of the electrode plate are completely consistent, and 0 indicates that the actual thickness of the electrode plate and the predicted thickness of the electrode plate are completely irrelevant, for example, if the actual thickness of the electrode plate of the ternary lithium battery is 120 μm, and the predicted thickness of the electrode plate predicted by the electrode thickness predictor is 118 μm, then the similarity of this electrode thickness group is = 1 - (|120-118|) / 120 = 0.983, and thus, the similarity of a plurality of electrode thickness groups is calculated in the same way, and the arithmetic mean is calculated, for example, 0.98, to obtain the basic thickness confidence, which can reflect the overall consistency between the actual thickness of the electrode plate and the predicted thickness of the electrode plate predicted by the electrode thickness predictor, and the greater the basic thickness confidence, the stronger the overall consistency between the actual thickness of the electrode plate and the predicted thickness of the electrode plate, that is, the more reliable the data.

[0077] Finally, the thickness confidence and the discharge confidence that are more suitable for the actual scene are calculated according to the test quantity and the prediction accuracy, because the basic thickness confidence and the basic discharge confidence only reflect the overall consistency between the actual thickness of the electrode plate, the actual discharge slope and the predicted thickness of the electrode plate, the predicted discharge slope predicted by the model, without considering the influence of the test scale (test quantity) and the model reliability (prediction accuracy) on the confidence: the more the test quantity, the more sufficient the data coverage of the actual thickness of the electrode plate and the discharge slope, the more thoroughly the accidental error is diluted, and the stronger the representativeness of the data; the higher the model prediction accuracy, the closer the predicted thickness of the electrode plate and the predicted discharge slope output to the true value, and the higher the reference value of the basic confidence, so it is necessary to quantify the influence of the test quantity and the prediction accuracy, so that the final output thickness confidence and discharge confidence are more suitable for the actual test scene, and the objectivity and reliability of the confidence evaluation are improved.

[0078] Specifically, the "calculating the thickness confidence and the discharge confidence according to the test quantity and the prediction accuracy" comprises:

[0079] obtaining a test quantity of the test;

[0080] calculating a ratio of the test quantity and an average test quantity to obtain a first confidence correction coefficient;

[0081] obtaining an accuracy of the discharge slope prediction and the thickness prediction of the electrode piece, and calculating a mean value to obtain a second confidence correction coefficient;

[0082] correcting and calculating the basic thickness confidence and the basic discharge confidence according to the first confidence correction coefficient and the second confidence correction coefficient to obtain the thickness confidence and the discharge confidence.

[0083] In the embodiment of the application, firstly, a test quantity of the test is obtained, for example, the test quantity of the test is 5, which indicates that there are 5 measured thicknesses of the electrode piece and 5 measured discharge slopes. The test quantity can reflect the representativeness of the data. Generally, the more the test quantity is, the wider the electrode piece area covered is, the more detailed the slope segmentation is, the higher the degree of dilution of accidental errors is, and the stronger the data reliability is.

[0084] Secondly, a ratio of the test quantity and an average test quantity is calculated to obtain a first confidence correction coefficient, wherein the first confidence correction coefficient = test quantity / average test quantity. The average test quantity refers to an average test quantity accumulated historically, which can be used as a benchmark value for measuring the sufficiency of the test data. If the first confidence correction coefficient is greater than 1, it indicates that the test quantity is more than the historical average level, the data coverage is more comprehensive, and the accidental errors are diluted more sufficiently. Therefore, the basic confidence can be positively improved. If the first confidence correction coefficient is less than 1, it indicates that the test quantity is less than the historical average level, the data representativeness is insufficient, and the basic confidence needs to be reduced to reflect the deficiency of the data sufficiency, so that the confidence result is more consistent with the sample size difference of the actual test. For example, if the average test quantity is 6 and the test quantity is 5, the first confidence correction coefficient = 5 / 6 = 0.83.

[0085] Thirdly, an accuracy of the discharge slope prediction of the discharge slope predictor and an accuracy of the thickness prediction of the electrode piece of the thickness predictor are obtained, and a mean value is calculated to obtain a second confidence correction coefficient. For example, the prediction accuracy of the discharge slope predictor and the prediction accuracy of the thickness predictor of the electrode piece are tested by an independent test set, respectively. For example, the prediction accuracy of the discharge slope predictor is 96%, and the prediction accuracy of the thickness predictor of the electrode piece is 94%. Then, the second confidence correction coefficient = (96% + 94%) / 2 = 95%. The greater the second confidence correction coefficient is, the higher the confidence of the multiple predicted thicknesses of the electrode piece and the predicted discharge slopes output by the discharge slope predictor and the thickness predictor of the electrode piece is, and the higher the basic confidence obtained by the calculation is.

[0086] Finally, the basic thickness confidence and the basic discharge confidence are calculated according to the first confidence correction coefficient and the second confidence correction coefficient to obtain the thickness confidence and the discharge confidence, wherein the thickness confidence = the basic thickness confidence * the first confidence correction coefficient * the second confidence correction coefficient, and the discharge confidence = the basic discharge confidence * the first confidence correction coefficient * the second confidence correction coefficient. The thickness confidence and the discharge confidence comprehensively consider the data consistency, the data sufficiency and the model reliability. The greater the thickness confidence and the discharge confidence are, the more reliable the measured data is, which can be used as a basis for evaluating the performance of the pole piece. If the thickness confidence and the discharge confidence are too low, the measured data is unreliable, and the model needs to be retested or optimized. For example, if the basic thickness confidence is 0.98, the basic discharge confidence is 0.96, the first confidence correction coefficient is 0.83, and the second confidence correction coefficient is 95%, then the thickness confidence = 0.98 * 0.83 * 95% = 0.77, and the discharge confidence = 0.96 * 0.83 * 95% = 0.76.

[0087] In summary, compared with the prior art, the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses are randomly combined in the present application, the thickness confidence is calculated and obtained, the plurality of discharge slopes and the plurality of predicted discharge slopes are randomly combined, the discharge confidence is calculated and obtained, and the confidence is calculated according to the test number and the prediction accuracy. In this way, the consistency of the measured data and the predicted data is converted into a quantifiable reliability index, and the influence of the test scale and the model reliability on the confidence is fully considered. The final output thickness confidence and discharge confidence provide a clear credibility label for the performance test results of the pole piece, and solve the pain point that the referenceability of the traditional test cannot be quantified.

[0088] S40: According to the thickness confidence and the discharge confidence, the performance test results of the pole piece of the ternary lithium battery are calculated by combining the plurality of pole piece thicknesses and the plurality of discharge slopes.

[0089] The physical properties (thickness) of the pole piece, the electrochemical performance (discharge performance) and the data reliability all affect the performance test results of the pole piece of the ternary lithium battery, so a systematic method is needed to integrate the three into the evaluation system to obtain the performance test results that fully reflect the true quality level of the pole piece.

[0090] To solve the above problems, the performance test results of the pole piece of the ternary lithium battery are calculated according to the thickness confidence and the discharge confidence by combining the plurality of pole piece thicknesses and the plurality of discharge slopes.

[0091] Specifically, step S40 in the method includes:

[0092] a thickness performance parameter is calculated by combining the average of the similarities of the plurality of thicknesses and the thickness confidence;

[0093] a discharge performance parameter is calculated by combining the average of the similarities of the plurality of discharge slopes and the discharge confidence;

[0094] a performance test result is calculated according to the thickness performance parameter and the discharge performance parameter.

[0095] In the embodiments of the present application, first, the average of the similarities of the plurality of thicknesses and the standard thickness of the pole piece is calculated, and the thickness performance parameter is calculated by combining the thickness confidence, wherein the standard thickness of the pole piece refers to the ideal thickness of the pole piece meeting the design requirements, such as 120 μm±5 μm for the positive electrode and 200 μm±8 μm for the negative electrode, which is usually pre-set by the process standard or performance target and is the benchmark for evaluating whether the physical characteristics of the pole piece meet the standard, and the thickness performance parameter=similarity average*thickness confidence. For example, the similarities of the plurality of thicknesses and the standard thickness of the pole piece are calculated, and the arithmetic average of the plurality of similarities is calculated, wherein the similarity can be calculated by the following formula:

[0096]

[0097] For example, the standard thickness of the pole piece is 120 μm, the allowable deviation is ±5 μm, and the five measured values are 122 μm, 119 μm, 124 μm, 118 μm, and 123 μm, respectively, and the similarities are 1−2 / 5=0.6, 1−1 / 5=0.8, 1−4 / 5=0.2, 1−2 / 5=0.6, and 1−3 / 5=0.4, respectively, the average of the similarities is (0.6+0.8+0.2+0.6+0.4) / 5=0.52, and if the thickness confidence is 0.77, the thickness performance parameter=0.52*0.77=0.40. The thickness performance parameter reflects the degree of fit between the thickness of the pole piece and the standard thickness of the pole piece, and also reflects the reliability of the measured data through the thickness confidence. Even if the average of the similarities of the thickness of the pole piece and the standard thickness of the pole piece is very high, if the thickness confidence is low (such as insufficient test quantity), the final thickness performance parameter will also be reduced, avoiding overestimating the value of unreliable data.

[0098] ​Secondly, the similarity average of the plurality of discharge slopes and the standard discharge slope is calculated, and a discharge performance parameter is calculated in combination with the discharge confidence, wherein the standard discharge slope refers to an ideal discharge slope meeting the performance requirements, which is determined by the battery design target (such as energy density, cycle life) and is a benchmark for evaluating electrochemical performance, and the discharge performance parameter = similarity average * discharge confidence. Exemplarily, the similarity average of the plurality of discharge slopes and the standard discharge slope is calculated according to the same logic and method as the foregoing steps, for example, if the standard discharge slope is -0.005 V / mAh, the allowable deviation is ±0.0005 V / mAh, and the five measured discharge slopes are -0.0052 V / mAh, -0.0049 V / mAh, -0.0052 V / mAh, -0.0048 V / mAh, and -0.0051 V / mAh, respectively, then the similarities are calculated as 1-0.0002 / 0.0005=0.6, 1-0.0001 / 0.0005=0.8, 1-0.0002 / 0.0005=0.6, 1-0.0002 / 0.0005=0.6, and 1-0.0001 / 0.0005=0.8, respectively, the similarity average is (0.6+0.8+0.6+0.6+0.8) / 5=0.68, and if the discharge confidence is 0.76, then the discharge performance parameter = 0.68*0.76=0.52. The discharge performance parameter comprehensively reflects the degree of coincidence between the electrode electrochemical performance and the standard value and the data reliability, and avoids misjudgment caused by high similarity but low confidence (such as inaccurate model prediction).

[0099] Finally, the performance test result is calculated according to the thickness performance parameter and the discharge performance parameter, wherein the performance test result = w1*thickness performance parameter + w2*discharge performance parameter, w1 and w2 are the influence weights of the thickness performance parameter and the discharge performance parameter on the overall performance of the battery, w1+w2=1, and a person skilled in the art can dynamically set them according to the actual situation, for example, w1=0.6 and w2=0.4. Exemplarily, if the thickness performance parameter is 0.40, the discharge performance parameter is 0.52, w1=0.6, and w2=0.4, then the performance test result = 0.6*0.40+0.4*0.52=0.448 at this time. The performance test result is a quantitative value of 0% to 100%, and the higher the value, the better the overall performance of the electrode, which meets the physical standard (thickness) and has good electrochemical performance (discharge performance) and reliable test data. A fixed threshold value can be set according to this to determine whether the performance of the ternary lithium battery electrode meets the standard, for example, the performance test result threshold is set to 70%, and the performance of the ternary lithium battery electrode below the performance test result threshold does not meet the standard and needs to be checked for thickness unevenness or abnormal reaction.

[0100] In summary, compared with the prior art, the performance test result of the ternary lithium battery pole piece is obtained according to the thickness confidence and the discharge confidence, in combination with the plurality of pole piece thicknesses and the plurality of discharge slopes. In this way, the influence of thickness, discharge performance, and data reliability on the performance of the ternary lithium battery pole piece is quantified, the performance evaluation is both comprehensive and reliable, and a direct landing quantitative basis is provided for pole piece quality grading and process optimization.

[0101] In summary, the embodiments of the present application have at least the following technical effects:

[0102] Compared with the prior art, the present application first performs multi-point thickness testing on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and tests and obtains a plurality of discharge slopes of the ternary lithium battery pole piece. In this way, by collecting representative and stable data, a reliable basis is provided for subsequent machine learning prediction and confidence analysis.

[0103] Secondly, the present application respectively predicts the discharge slope according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and predicts the pole piece thickness according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses. In this way, the cross-prediction of pole piece thickness→discharge slope and discharge slope→pole piece thickness is realized through the machine learning model, the data consistency is verified through bidirectional mapping, and a prediction benchmark is provided for subsequent confidence calculation.

[0104] Thirdly, the present application randomly combines the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to calculate a thickness confidence, and randomly combines the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate a discharge confidence, wherein the confidence is calculated according to the test quantity and the prediction accuracy. In this way, the consistency of the measured data and the predicted data is converted into a quantifiable reliability index, and the influence of the test scale and the model reliability on the confidence is fully considered. The finally output thickness confidence and discharge confidence provide a clear credibility label for the pole piece performance test result, and solve the pain point that the referenceability of traditional testing cannot be quantified.

[0105] Finally, the present application obtains the performance test result of the ternary lithium battery pole piece according to the thickness confidence and the discharge confidence, in combination with the plurality of pole piece thicknesses and the plurality of discharge slopes. In this way, the influence of thickness, discharge performance, and data reliability on the performance of the ternary lithium battery pole piece is quantified, the performance evaluation is both comprehensive and reliable, and a direct landing quantitative basis is provided for pole piece quality grading and process optimization.

[0106] Through the above technical solution, this application obtains multiple electrode thicknesses and corresponding discharge slopes by conducting multi-point tests on the ternary lithium battery electrode, and realizes bidirectional prediction of electrode thickness → predicted discharge slope and discharge slope → predicted electrode thickness based on the machine learning model, forming a cross-validation system of measured data and predicted data, and then calculates the basic confidence by randomly combining the measured and predicted data, and performs double corrections based on the number of tests (data sufficiency) and prediction accuracy (model reliability), accurately quantifying the thickness confidence and discharge confidence, and finally integrating multi-dimensional indicators to calculate the comprehensive performance test results of the electrode. In this way, the accidental errors of traditional single-point detection are effectively avoided through multi-point data collection and bidirectional prediction, and the representativeness and relevance of the data are improved. The traceable evaluation of the reliability of the test results is realized through confidence quantification, which improves the representativeness and credibility of the performance test data of the ternary lithium battery electrode, and provides high-precision, quantifiable and reliable technical support for the quality control of the ternary lithium battery electrode, production process optimization and battery performance improvement.

[0107] Example 2, as Figure 2 As shown, based on the same inventive concept as the method for testing the performance of a ternary lithium battery pole piece by integrating machine learning provided in Example 1, an embodiment of the present invention also provides a system for testing the performance of a ternary lithium battery pole piece by integrating machine learning, comprising:

[0108] The thickness testing module 11 is used to perform multi-point thickness testing on the ternary lithium battery electrode to obtain multiple electrode thicknesses and to test and obtain multiple discharge slopes of the ternary lithium battery electrode;

[0109] A parameter prediction module 12 is used to predict the discharge slope according to the plurality of electrode thicknesses to obtain a plurality of predicted discharge slopes, and to predict the electrode thickness according to the plurality of discharge slopes to obtain a plurality of predicted electrode thicknesses;

[0110] A confidence analysis module 13 is configured to randomly combine the plurality of electrode thicknesses and the plurality of predicted electrode thicknesses to calculate thickness confidence, and to randomly combine the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate discharge confidence, wherein the confidence is calculated based on the number of tests and the prediction accuracy;

[0111] The fusion output module 14 is used to calculate the performance test results of the ternary lithium battery electrode according to the thickness confidence and discharge confidence, combined with multiple electrode thicknesses and multiple discharge slopes.

[0112] The thickness testing module 11 is specifically used for:

[0113] Get the number of tests;

[0114] Randomly select the test quantity of test positions on the ternary lithium battery pole piece, perform multi-point thickness testing, and obtain a plurality of pole piece thicknesses;

[0115] Perform constant current discharge testing on the ternary lithium battery to obtain a discharge curve at a medium SOC;

[0116] According to the test quantity, divide the discharge curve, and obtain the slopes of the divided plurality of discharge curve segments as a plurality of discharge slopes.

[0117] The parameter prediction module 12 is specifically configured to:

[0118] Call a discharge slope predictor;

[0119] Input the plurality of pole piece thicknesses into the discharge slope predictor respectively, and obtain a plurality of predicted discharge slopes through prediction output;

[0120] Call a pole piece thickness predictor;

[0121] Input the plurality of discharge slopes into the pole piece thickness predictor respectively, and obtain a plurality of predicted pole piece thicknesses through prediction output.

[0122] Specifically, the "calling a discharge slope predictor" includes:

[0123] According to historical test data of the ternary lithium battery pole piece, a sample pole piece thickness set and a constant current discharge curve slope of the ternary lithium battery under different sample pole piece thicknesses are collected, and a sample discharge slope set is labeled and obtained;

[0124] Construct a discharge slope predictor based on machine learning;

[0125] The sample pole piece thickness set and the sample discharge slope set are used to supervise the training of the discharge slope predictor, and the training is completed after the test converges;

[0126] The converged discharge slope predictor is configured in a cloud server.

[0127] The confidence analysis module 13 is specifically configured to:

[0128] Randomly combine the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to obtain a plurality of pole piece thickness groups, calculate the similarity respectively and calculate the mean value to obtain a basic thickness confidence;

[0129] Randomly combine the plurality of discharge slopes and the plurality of predicted discharge slopes to obtain a plurality of discharge slope groups, calculate the similarity respectively and calculate the mean value to obtain a basic slope confidence;

[0130] According to the test quantity and the prediction accuracy, a thickness confidence and a discharge confidence are calculated and obtained.

[0131] Specifically, the "calculating the thickness confidence and the discharge confidence according to the test quantity and the prediction accuracy" comprises:

[0132] obtaining the test quantity of the test;

[0133] calculating the ratio of the test quantity and the average test quantity to obtain a first confidence correction coefficient;

[0134] obtaining the accuracy of the discharge slope prediction and the thickness prediction of the pole piece, and calculating the average to obtain a second confidence correction coefficient;

[0135] correcting and calculating the basic thickness confidence and the basic discharge confidence according to the first confidence correction coefficient and the second confidence correction coefficient to obtain the thickness confidence and the discharge confidence.

[0136] The fusion output module 14 is specifically configured to:

[0137] calculating the average of the similarity of the plurality of pole piece thicknesses and the standard pole piece thickness, and combining the thickness confidence to obtain a thickness performance parameter;

[0138] calculating the average of the similarity of the plurality of discharge slopes and the standard discharge slope, and combining the discharge confidence to obtain a discharge performance parameter;

[0139] calculating the performance test result according to the thickness performance parameter and the discharge performance parameter.

[0140] In summary, the embodiments of the present application have at least the following technical effects:

[0141] Compared with the prior art, firstly, the thickness test module is used to perform multi-point thickness testing on the ternary lithium battery pole piece to obtain a plurality of pole piece thicknesses, and a plurality of discharge slopes of the ternary lithium battery pole piece are tested and obtained, so that representative and stable data are collected to provide a reliable basis for subsequent machine learning prediction and confidence analysis. Secondly, the parameter prediction module is used to respectively perform discharge slope prediction according to the plurality of pole piece thicknesses to obtain a plurality of predicted discharge slopes, and to perform pole piece thickness prediction according to the plurality of discharge slopes to obtain a plurality of predicted pole piece thicknesses, so that bidirectional prediction of pole piece thickness→predicted discharge slope and discharge slope→predicted pole piece thickness is realized through a machine learning model, and data consistency is verified through bidirectional mapping to provide a prediction benchmark for subsequent confidence calculation. Thirdly, the confidence analysis module is used to randomly combine the plurality of pole piece thicknesses and the plurality of predicted pole piece thicknesses to calculate a thickness confidence, and to randomly combine the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate a discharge confidence, so that the consistency of the measured data and the predicted data is converted into a quantifiable reliability index, and the influence of the test scale and the model reliability on the confidence is fully considered, and finally the thickness confidence and the discharge confidence are output, so that a clear confidence label is provided for the pole piece performance test result, and the pain point that the referenceability of the traditional test cannot be quantified is solved. Finally, the fusion output module is used to calculate a performance test result of the ternary lithium battery pole piece according to the thickness confidence and the discharge confidence in combination with the plurality of pole piece thicknesses and the plurality of discharge slopes, so that the influence of the thickness, the discharge performance, and the data reliability on the performance of the ternary lithium battery pole piece is quantified, the performance evaluation is comprehensive and reliable, and a quantifiable basis is provided for the pole piece quality grading and process optimization. In this way, the accidental error of the traditional single-point detection is effectively avoided through multi-measurement-point data collection and bidirectional prediction, the data representativeness and relevance are improved, the traceable evaluation of the reliability of the test result is realized through confidence quantification, the representativeness and the reliability of the performance test data of the ternary lithium battery pole piece are improved, and high-precision, quantifiable, and reliable technical support is provided for the quality control, the production process optimization, and the performance improvement of the ternary lithium battery pole piece.

[0142] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0143] Those skilled in the art will appreciate that embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0144] The present application is described in reference to the flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the flowchart block or blocks.

[0145] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the flowchart block or blocks.

[0146] The computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the flowchart block or blocks.

[0147] Although preferred embodiments of the application have been described, those skilled in the art will recognize that additional modifications and variations are possible without departing from the spirit and scope of the application.

[0148] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the application, modifications and variations of the application can be made by those skilled in the art upon celebrating the above teaching and once the application starts. Accordingly, the application is to be understood as including all such modifications and variations as falling within the scope of the application and equivalents of the same.

Claims

1. A method for testing the performance of ternary lithium battery pole pieces by integrating machine learning, characterized in that: include: Perform multi-point thickness testing on the ternary lithium battery electrode to obtain multiple electrode thicknesses, and test and obtain multiple discharge slopes of the ternary lithium battery electrode; Predicting discharge slopes according to multiple electrode thicknesses to obtain multiple predicted discharge slopes, and predicting electrode thicknesses according to multiple discharge slopes to obtain multiple predicted electrode thicknesses; Randomly combining the plurality of electrode thicknesses and the plurality of predicted electrode thicknesses to calculate thickness confidence, and randomly combining the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate discharge confidence, wherein the confidence is calculated based on the number of tests and the prediction accuracy; According to the thickness confidence and discharge confidence, the performance test results of the ternary lithium battery electrode are obtained by combining multiple electrode thicknesses and multiple discharge slopes.

2. The ternary lithium battery pole piece performance testing method integrating machine learning according to claim 1 is characterized in that: Perform multi-point thickness testing on the ternary lithium battery electrode to obtain multiple electrode thicknesses, and test and obtain multiple discharge slopes of the ternary lithium battery electrode, including: Get the number of tests; Randomly select the test number of test positions on the ternary lithium battery electrode, perform multi-point thickness testing, and obtain multiple electrode thicknesses; Perform a constant current discharge test on the ternary lithium battery to obtain the discharge curve at medium SOC; The discharge curve is divided according to the test quantity, and slopes of the divided multiple discharge curve segments are obtained as multiple discharge slopes.

3. The ternary lithium battery pole piece performance testing method integrating machine learning according to claim 1 is characterized in that: Discharge slopes are predicted based on multiple electrode thicknesses to obtain multiple predicted discharge slopes. Electrode thicknesses are predicted based on multiple discharge slopes to obtain multiple predicted electrode thicknesses, including: Call the discharge slope predictor; Inputting the plurality of electrode thicknesses into the discharge slope predictor respectively, and outputting a prediction to obtain a plurality of predicted discharge slopes; Call the pole piece thickness predictor; The multiple discharge slopes are respectively input into the electrode thickness predictor, and the prediction output obtains multiple predicted electrode thicknesses.

4. The method for testing the performance of a ternary lithium battery electrode piece by integrating machine learning according to claim 3, characterized in that: Call the discharge slope predictor, including: Based on the historical test data of ternary lithium battery electrodes, a set of sample electrode thicknesses and the slopes of the constant current discharge curves of ternary lithium batteries under different sample electrode thicknesses are collected, and the sample discharge slope set is marked; Build a machine learning-based discharge slope predictor; Using the sample electrode thickness set and the sample discharge slope set, supervised training is performed on the discharge slope predictor, and the training is completed after test convergence; The converged discharge slope predictor is configured on the cloud server.

5. The method for testing the performance of a ternary lithium battery electrode piece by integrating machine learning according to claim 1, characterized in that: Randomly combining the plurality of electrode thicknesses and the plurality of predicted electrode thicknesses to calculate thickness confidence, and randomly combining the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate discharge confidence, including: Randomly combining the multiple pole piece thicknesses and the multiple predicted pole piece thicknesses to obtain multiple pole piece thickness groups, calculating similarities and averages for each group, and obtaining basic thickness confidence; Randomly combining the multiple discharge slopes and the multiple predicted discharge slopes to obtain multiple discharge slope groups, calculating similarities and averages for each of the groups to obtain basic slope confidences; According to the number of tests and the prediction accuracy, the thickness confidence and discharge confidence are calculated.

6. The method for testing the performance of a ternary lithium battery electrode piece by integrating machine learning according to claim 5, characterized in that: Based on the number of tests and the prediction accuracy, the thickness confidence and discharge confidence are calculated, including: Get the number of tests to be tested; Calculating a ratio of the number of tests to an average number of tests to obtain a first confidence correction coefficient; Obtain the accuracy of the discharge slope prediction and the electrode thickness prediction, calculate the mean and obtain the second confidence correction coefficient; The basic thickness confidence and basic discharge confidence are corrected and calculated according to the first confidence correction coefficient and the second confidence correction coefficient to obtain thickness confidence and discharge confidence.

7. The method for testing the performance of a ternary lithium battery electrode piece by integrating machine learning according to claim 1, characterized in that: According to the thickness confidence and discharge confidence, the performance test results of the ternary lithium battery electrode are obtained by combining multiple electrode thicknesses and multiple discharge slopes, including: Calculating the average similarity between the thicknesses of multiple pole pieces and the thickness of a standard pole piece, and combining the thickness confidence to obtain the thickness performance parameter; Calculating the average similarity between the plurality of discharge slopes and the standard discharge slope, and combining the discharge confidence to obtain the discharge performance parameter; The performance test results are obtained by calculation based on the thickness performance parameters and the discharge performance parameters.

8. The ternary lithium battery pole piece performance test system integrating machine learning is characterized by: Used to perform the method according to any one of claims 1 to 7, comprising: Thickness testing module, used to perform multi-point thickness testing on ternary lithium battery pole pieces, obtain multiple pole piece thicknesses, and test and obtain multiple discharge slopes of ternary lithium battery pole pieces; A parameter prediction module is used to predict the discharge slope according to the thickness of multiple electrode sheets to obtain multiple predicted discharge slopes, and to predict the electrode thickness according to the multiple discharge slopes to obtain multiple predicted electrode thicknesses; A confidence analysis module, configured to randomly combine the plurality of electrode thicknesses and the plurality of predicted electrode thicknesses to calculate thickness confidence, and to randomly combine the plurality of discharge slopes and the plurality of predicted discharge slopes to calculate discharge confidence, wherein the confidence is calculated based on the number of tests and the prediction accuracy; The fusion output module is used to calculate the performance test results of the ternary lithium battery electrode according to the thickness confidence and discharge confidence, combined with multiple electrode thicknesses and multiple discharge slopes.

Images