Prediction method, system and equipment for coating quality of battery pole piece and storage medium

Abstract

The invention relates to the technical field of coating quality prediction, and discloses a battery pole piece coating quality prediction method, system and device and a storage medium, and the method comprises the steps: obtaining the coating time sequence data of a battery pole piece, calculating the similarity between the coating time sequence data and the clustering centroid corresponding to each preset sample type, and obtaining the coating quality of the battery pole piece according to the similarity. Obtaining a sample type of the coating time sequence data; selecting a time sequence prediction model corresponding to the sample type from a pre-established time sequence prediction model set, and inputting the coating time sequence data into the time sequence prediction model to obtain coating time sequence prediction data; and determining the coating quality of the battery pole piece according to the coating time sequence prediction data. According to the method, the high-quality sample data set is established, and the classified time sequence prediction model is adopted to predict the coating data, so that the accuracy of a coating quality data prediction result is effectively improved, and the quality and efficiency of battery pole piece coating production are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coating quality prediction, and particularly to a method, a system, a device and a storage medium for predicting the coating quality of battery electrode sheets. Background Art

[0002] The battery electrode sheet is a core component of the battery. During the production of the battery electrode sheet, processes such as slurry preparation, coating, rolling and shearing are required. Among them, coating refers to the process of using a coating device to coat a fluid coating made of a mixture of electrode materials on a substrate during the production of the electrode, and then drying and forming. Coating is a key process in the production of battery electrode sheets. The coating quality of battery electrode sheets has an important impact on battery performance. The coating quality directly affects the capacity, internal resistance, cycle life and safety of the battery. Therefore, during the production process of battery electrode sheet coating, various parameters need to be adjusted according to the quality of the electrode sheet to ensure the coating quality of the battery electrode sheet.

[0003] Currently, for battery electrode sheets, the coating quality is usually determined by means of coating quality detection or coating quality prediction. Although the coating quality detection method can accurately judge the coating quality of battery electrode sheets, the current coating quality detection devices usually have a long time lag, resulting in abnormal coating not being detected in time, and the parameters of the coating device cannot be adjusted in time, thus affecting the quality and efficiency of electrode coating production; while the existing coating quality prediction methods usually use multi-factor prediction. Therefore, a large number of factors affecting the coating quality of electrode sheets need to be collected and screened. The collection and selection of key influencing factors will directly affect the accuracy and stability of the prediction results. At present, there is no good method for collecting and selecting key influencing factors, and it is difficult to collect key influencing factors, resulting in difficulty in establishing a conventional multi-factor prediction model and low prediction result accuracy. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a method, a system, a device and a storage medium for predicting the coating quality of battery electrode sheets, so as to solve the problems of low prediction accuracy and untimely detection of the existing methods, and achieve the technical effect of timely and accurate prediction of the coating quality of battery electrode sheets, thereby improving the production quality and efficiency.

[0005] In a first aspect, the present invention provides a method for predicting the coating quality of battery electrode sheets, the method comprising:

[0006] Obtain the coating time series data of the battery electrode sheet, calculate the similarity between the coating time series data and the clustering centroids corresponding to each preset sample type, and obtain the sample type of the coating time series data according to the similarity. The clustering centroids are obtained by performing clustering analysis on a preset sample data set using a clustering algorithm. The sample data in the sample data set includes coating quality data;

[0007] Select the time series prediction model corresponding to the sample type from a pre-established set of time series prediction models, and input the coating time series data into the time series prediction model to obtain coating time series prediction data, where the set of time series prediction models includes a number of the time series prediction models, the time series prediction models correspond one-to-one with the sample types, and each of the time series prediction models is constructed using a neural network model and trained using the sample data set;

[0008] Determine the coating quality of the battery electrode sheet according to the coating time series prediction data.

[0009] Further, the steps for constructing the sample data set include:

[0010] Obtain the historical coating data of the battery electrode sheet, and preprocess the historical coating data to obtain a coating quality time series. The historical coating data includes coating quality data, sampling time, coating speed, and coating die adjustment identification;

[0011] Use the coating die adjustment identification that is the same as the preset identification value as the center point, and intercept the coating quality time series forward according to a preset length to obtain a number of coating quality time subsequences. The preset length is determined based on the length of the coating oven;

[0012] Equally divide the coating quality time subsequences to obtain first coating quality time series data and second coating quality time series data, and establish a sample data set according to the first coating quality time series data and the second coating quality time series data;

[0013] Divide the sample data set into a first sample subset and a second sample subset according to a preset ratio;

[0014] Use the K-means clustering algorithm based on dynamic time warping to perform clustering analysis on each of the first coating quality time series data in the first sample subset to obtain the clustering centroids corresponding to each sample type;

[0015] Use the dynamic time warping algorithm to calculate the first similarity between each of the first coating quality time series data in the second sample subset and each of the clustering centroids, and classify the second sample subset according to the first similarity to obtain a sample data subset corresponding to each sample type.

[0016] Further, the step of preprocessing the historical coating data to obtain a coating quality time series includes:

[0017] Perform standardization processing on the historical coating data to obtain first standardized coating data;

[0018] Calculate the sampling moving distance according to the sampling time and the coating speed, and update the first standardized coating data according to the sampling moving distance to obtain the second standardized coating data;

[0019] Update the coating quality data in the second standardized coating data according to a preset sampling frequency and the sampling moving distance, and establish a coating quality time series according to the updated coating quality data.

[0020] Further, the step of calculating the sampling moving distance according to the sampling time and the coating speed, and updating the first standardized coating data according to the sampling moving distance to obtain the second standardized coating data includes:

[0021] Calculate the sampling distance interval according to the sampling time and the coating speed, and establish a sampling distance interval sequence according to the sampling time;

[0022] Set the first sampling distance interval in the sampling distance interval sequence to zero, and perform a cumulative sum calculation on the sampling distance interval sequence after setting to zero to obtain a sampling moving distance sequence, which is composed of the sampling moving distances corresponding to each sampling time;

[0023] Replace the sampling time and the coating speed in the first standardized coating data with the corresponding sampling moving distance to obtain the second standardized coating data.

[0024] Further, the step of updating the coating quality data in the second standardized coating data according to a preset sampling frequency and the sampling moving distance includes:

[0025] Use the sampling moving distance as the dividend and the preset sampling frequency as the divisor to calculate the remainder between each sampling moving distance and the sampling frequency;

[0026] Judge whether the remainder is equal to the remainder threshold. If not, linearly interpolate and update the coating quality data corresponding to the sampling moving distance according to the sampling frequency and the sampling moving distance to obtain the updated coating quality data. If so, keep the coating quality data corresponding to the sampling moving distance unchanged;

[0027] Among them, the updated coating quality data is represented by the following formula:

[0028]

[0029] In the formula, m′ kdenote the coating quality data of the k-th update, f denote the sampling frequency, m k denote the k-th coating quality data, m k-1 denote the (k - 1)-th coating quality data, l k denote the k-th sampling moving distance, l k-1 denote the (k - 1)-th sampling moving distance.

[0030] Further, the steps of calculating the similarity between the coating time series data and the clustering centroids corresponding to each preset sample type, and obtaining the sample type of the coating time series data according to the similarity include:

[0031] Use the dynamic time warping algorithm to calculate the second similarity between the coating time series data and each of the clustering centroids;

[0032] According to the second similarity, select the clustering centroid with the highest similarity to the coating time series data from each of the clustering centroids as the membership clustering centroid;

[0033] Determine the sample type of the coating time series data according to the membership clustering centroid.

[0034] Further, each of the time series prediction models is trained using the corresponding subset of sample data according to the sample type, where the first coating quality time series data in the subset of sample data is sample data, and the second coating quality time series data in the subset of sample data is label data.

[0035] In a second aspect, the present invention provides a prediction system for the coating quality of battery electrodes, the system includes:

[0036] A sample classification module, configured to obtain the coating time series data of the battery electrode, calculate the similarity between the coating time series data and the clustering centroids corresponding to each preset sample type, and obtain the sample type of the coating time series data according to the similarity, the clustering centroids are obtained by performing clustering analysis on a preset sample data set using a clustering algorithm, and the sample data in the sample data set includes coating quality data;

[0037] A time series prediction module, configured to select the time series prediction model corresponding to the sample type from a pre-established set of time series prediction models, and input the coating time series data into the time series prediction model to obtain coating time series prediction data, where the set of time series prediction models includes several time series prediction models, the time series prediction models are in one-to-one correspondence with the sample types, and each of the time series prediction models is constructed using a neural network model and trained using the sample data set;

[0038] A quality analysis module for determining the coating quality of the battery electrode sheet according to the predicted coating timing data.

[0039] In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0040] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0041] The present invention provides a method, system, device, and storage medium for predicting the coating quality of battery electrode sheets. By establishing a high-quality sample data set, the present invention overcomes the instability of data collection in the actual coating process, improves the prediction accuracy of the trained timing prediction model, and further improves the accuracy of coating data prediction by establishing a set of classification prediction models. Through accurate coating prediction data, the present invention can provide timely and effective reference for the adjustment of coating production parameters, thus effectively solving the problem of lag in coating production quality adjustment, and further improving the quality and efficiency of battery electrode sheet coating production. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic flowchart of a method for predicting the coating quality of battery electrode sheets in an embodiment of the present invention;

[0043] Figure 2 is a schematic structural diagram of a system for predicting the coating quality of battery electrode sheets in an embodiment of the present invention;

[0044] Figure 3 is an internal structural diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Please refer to Figure 1 , a method for predicting the coating quality of battery electrode sheets proposed in the first embodiment of the present invention includes steps S10 to S30:

[0047] Step S10: Obtain the coating timing data of the battery electrode sheet, calculate the similarity between the coating timing data and the clustering centroids corresponding to each preset sample type, and obtain the sample type of the coating timing data according to the similarity. The clustering centroids are obtained by performing clustering analysis on the preset sample data set using a clustering algorithm. The sample data in the sample data set includes coating quality data;

[0048] Step S20: Select the timing prediction model corresponding to the sample type from the pre-established set of timing prediction models, and input the coating timing data into the timing prediction model to obtain the coating timing prediction data. The set of timing prediction models includes several timing prediction models, and the timing prediction models correspond to the sample types one by one. Each timing prediction model is constructed using a neural network model and trained using the sample data set;

[0049] Step S30: Determine the coating quality of the battery electrode sheet according to the coating timing prediction data.

[0050] The prediction method for the coating quality of the battery electrode sheet provided by the present invention is a classification prediction method. By establishing a set of timing prediction models, the quality of the coating data is predicted. The set of timing prediction models contains multiple timing prediction models, and each timing prediction model corresponds to a sample type. The sample type refers to the data type of the coating data. For example, by analyzing the historical data of the coating quality, it is determined that the coating data can be divided into N types. Then the sample types include the first sample type, the second sample type to the Nth sample type. Each sample type corresponds to a timing prediction model. These timing prediction models are constructed based on the neural network model, and each timing prediction model is trained using the data set of its corresponding sample type.

[0051] When establishing the set of timing prediction models, first classify the historical data of the coating quality, then construct the data sets corresponding to each sample type, and at the same time construct the timing prediction models corresponding to each sample type and train them using various data sets respectively to improve the prediction accuracy. During actual prediction, first classify the real-time collected coating timing data to determine the sample type to which the coating timing data belongs, then select the timing prediction model corresponding to the sample type, predict the change trend of the coating quality through this timing prediction model, and finally determine the coating quality of the battery electrode sheet according to the predicted coating timing data, so as to facilitate subsequent adjustment of the production parameters of the coating device according to the coating quality, thereby ensuring the production quality and efficiency of the battery electrode sheet.

[0052] Based on the above steps, in this embodiment, first, historical coating data of the battery electrode is obtained and classified to construct datasets of various types for subsequent training of the prediction model. When classifying the data, conventional data classification methods can be used, such as classifying the data through data analysis software or using data statistical analysis algorithms, and then constructing datasets for the classified data. Although this conventional method can construct classified datasets, in the actual coating process, factors such as changes in the die data of the coating device and changes in the acquisition frequency and speed of the coating data acquisition device will affect the collected coating data. Training the prediction model based on this data will inevitably affect the accuracy and stability of the prediction results. To improve the quality of the dataset and thus enhance the accuracy and robustness of the model, in a preferred embodiment, the present invention provides a method for constructing a dataset, and the specific steps include:

[0053] Obtain historical coating data of the battery electrode and preprocess the historical coating data to obtain a coating quality time series. The historical coating data includes coating quality data, sampling time, coating speed, and coating die adjustment identifier;

[0054] Use the coating die adjustment identifier that is the same as the preset identifier value as the center point, and intercept the coating quality time series forward according to a preset length to obtain a number of coating quality time subsequences. The preset length is determined based on the length of the coating oven;

[0055] Equally divide the coating quality time subsequences to obtain first coating quality time series data and second coating quality time series data, and establish a sample dataset based on the first coating quality time series data and the second coating quality time series data;

[0056] Divide the sample dataset into a first sample subset and a second sample subset according to a preset ratio;

[0057] Use the K-means clustering algorithm based on dynamic time warping to perform clustering analysis on each first coating quality time series data in the first sample subset to obtain the clustering centroids corresponding to each sample type;

[0058] Use the dynamic time warping algorithm to calculate the first similarity between each first coating quality time series data in the second sample subset and each clustering centroid, and classify the second sample subset according to the first similarity to obtain the sample data subsets corresponding to each sample type.

[0059] In this embodiment, first, historical coating data of the battery electrode sheet is obtained. The historical coating data includes coating quality data, sampling time, coating speed, and coating die adjustment identifier. Among them, the coating quality data refers to the data characterizing the coating quality, such as the coating density, coating thickness, or coating weight per unit area, which are data reflecting the quality of the coating process, or the average quality data of the overall area of the battery electrode sheet. These data can be collected by a detection device. According to the selected coating quality data, the detection device can be a surface density meter, a thickness detector, or a weight detector. When collecting data, the detection device will collect data by partitioning according to the coating area. Here, the coating area can be divided based on the detection range of the detection device or according to the size of the battery electrode sheet, and no excessive restrictions are imposed here. Moreover, to ensure the quality of the sample data set, the selected historical coating data should cover all types of quality change trends, so as to ensure the prediction effect of the subsequent trained time series prediction model on various types of data.

[0060] The historical coating data in this embodiment can be extracted from production records or historical databases. Data preprocessing is performed on these historical coating data, and the preprocessing methods can adopt conventional methods such as normalization, outlier removal, and missing value filling. Then, a time series of coating quality data is established based on time. In the actual coating process, the die data of the coating device will change, and the acquisition frequency and speed of the detection device will also change, resulting in unstable or unevenly collected data. To improve the data quality, a preferred data preprocessing method is provided in this embodiment. The specific steps include:

[0061] Perform standardization processing on the historical coating data to obtain the first standardized coating data;

[0062] Calculate the sampling moving distance according to the sampling time and coating speed, and update the first standardized coating data according to the sampling moving distance to obtain the second standardized coating data;

[0063] Update the coating quality data in the second standardized coating data according to the preset sampling frequency and sampling moving distance, and establish a coating quality time series according to the updated coating quality data.

[0064] In this embodiment, first, standardization processing is performed on the historical coating data. Specifically, standardization is performed on the coating quality data, and its standardization formula is:

[0065] m = (M - mean) / std

[0066] In the formula, M represents the original coating quality data, mean represents the data mean, std represents the data standard deviation, and m represents the standardized coating quality data.

[0067] Then, based on the standardized historical coating data, the first standardized coating data is obtained:

[0068]

[0069] where m i represents the coating quality data at the i-th moment, t i represents the sampling time at the i-th moment, v i represents the coating speed at the i-th moment, and signal i represents the coating die adjustment identifier at the i-th moment. The coating die adjustment identifier is used to indicate whether the die data of the coating device has changed. If the die data of the coating device changes due to the device itself or manual adjustment, then set signal i at this moment to 1. If there is no change at this moment, set this identifier to zero.

[0070] In the actual coating process, due to the change in the data acquisition frequency of the detection device and the change in the speed of the electrode plate passing through the detection device, both will cause the instability and non-uniformity of the acquired data. Therefore, it is necessary to perform interval processing on data sampling, convert the sampling time into a sampling distance interval, that is, calculate the sampling moving distance through the sampling time and the coating speed, and update the first standardized coating data according to the sampling moving distance. The specific steps include:

[0071] Calculate the sampling distance interval according to the sampling time and the coating speed, and establish a sampling distance interval sequence according to the sampling time;

[0072] Set the first sampling distance interval in the sampling distance interval sequence to zero, and perform cumulative sum calculation on the sampling distance interval sequence after setting it to zero to obtain a sampling moving distance sequence. The sampling moving distance sequence is composed of the sampling moving distances corresponding to each sampling time;

[0073] Replace the sampling time and the coating speed in the first standardized coating data with the corresponding sampling moving distances to obtain the second standardized coating data.

[0074] In this embodiment, the distance interval between two samplings is calculated through the sampling time and the coating speed. The sampling distance interval can be expressed as: s i =(t i -t i-1 )×v i , and then establish a sampling distance interval sequence [s0…s i …s n according to the sampling time. In this sequence, s0 represents the sampling distance interval corresponding to the initial sampling time, so set it to zero, that is, s0 = 0.

[0075] Then, perform a cumulative sum calculation on the zeroed sampling distance interval sequence. The cumulative sum calculation, that is, cumsum, is used to calculate the cumulative sum of each element in the sequence. Its function is to add the current element to the sum of all previous elements to obtain a new sequence, that is: Each element in this sequence represents the sampling moving distance of each sampling time relative to the initial sampling. Then, replace the sampling moving distance with the corresponding sampling time and coating speed in the first normalized coating data to obtain the second normalized coating data:

[0076]

[0077] For the second normalized coating data, unify the data acquisition frequency through linear interpolation. The specific steps include:

[0078] Take the sampling moving distance as the dividend and the preset sampling frequency as the divisor, and calculate the remainder between each sampling moving distance and the sampling frequency;

[0079] Judge whether the remainder is equal to the remainder threshold. If not, linearly interpolate and update the coating quality data corresponding to the sampling moving distance according to the sampling frequency and the sampling moving distance to obtain the updated coating quality data. If so, keep the coating quality data corresponding to the sampling moving distance unchanged.

[0080] In this embodiment, based on the conditions of the data acquisition device, the sampling frequency f is preset. For the convenience of description, Denoted by l n That is, l i Represents the i-th moment, that is, the i-th sampling moving distance. Then, sequentially judge whether the remainder between each sampling moving distance and the sampling frequency is zero, that is, judge whether l k mod f is zero. If l k mod f = 0, then keep the coating quality data unchanged, that is: m′ k = m k . If l k mod f ≠ 0, then linearly interpolate and update the coating quality data, that is:

[0081]

[0082] In the formula, m′ k Represents the k-th updated coating quality data, f represents the sampling frequency, m k Represents the k-th coating quality data, m k-1 Represents the (k - 1)-th coating quality data, l k Represents the k-th sampling moving distance, l k-1 Represents the (k - 1)-th sampling moving distance.

[0083] The second standardized coating data obtained at this time can be expressed as:

[0084]

[0085] It can be seen that compared with the previous second standardized coating data, the coating quality data is updated in the second standardized coating data at this time. Then, the coating quality data is extracted from the second standardized coating data, and a coating quality time series [m′0…m′ n is established.

[0086] It should be noted here that in this embodiment, the reason for further processing the coating quality data is to overcome the problems of data instability and unevenness caused by changes in the sampling frequency and speed during the actual coating process, so as to improve the data quality. In fact, a coating quality time series can also be established based on the first standardized coating data or the second standardized coating data. That is, the present invention provides a preferred data processing method rather than a specific limitation, and can be flexibly selected according to the prediction accuracy requirements in actual applications.

[0087] Then, based on the updated second standardized coating data and the coating quality time series, a sample data set is established. Specifically, centered on each data with signal = 1, the coating quality time series is intercepted forward. For the convenience of description, here the coating quality time series is represented by [m0…m n . Assuming that the interception length is 2L, after forward interception, several coating quality time subsequences [m0…m 2L-1 can be obtained, where L is determined by the length of the coating oven. Assuming that the length of the coating oven is a, preferably L can be set to a / 5. Of course, L can also be set to other multiples of the oven length. It should be noted here that the purpose of intercepting according to the oven length is to make the coating quality data in each coating quality time subsequence be approximately at the same moment of the coating oven, so as to improve the accuracy of subsequent prediction.

[0088] Then, each coating quality time subsequence is equally divided into two equal-length sequences, namely the first coating quality time series data [ma0…ma L-1 and the second coating quality time series data [mb0…mb L-1 . Let the first coating quality time series data be represented by X and the second coating quality time series data be represented by Y. Then, (X, Y) is used as a training sample to establish a sample data set, and if there are outliers in the time series data, the training sample is deleted.

[0089] Based on the sample data set below, sample subsets of different data types are established. Specifically, according to a preset ratio, the sample data set is divided into a first sample subset and a second sample subset. The preset ratio here can be 3:7 or 2:8. Of course, other division ratios can also be selected. The purpose of dividing the sample data set in this embodiment is to classify the second sample subset through the clustering analysis of the first sample subset. Therefore, the division ratio can be flexibly set according to the requirements of classification accuracy or calculation efficiency, and no specific limitation is made here.

[0090] In this embodiment, clustering analysis is performed on the first sample subset, that is, clustering analysis is performed on the first coating quality time series data X in the first sample subset. The clustering analysis algorithm can adopt a conventional clustering analysis algorithm. In this embodiment, the K-means clustering algorithm based on dynamic time warping is preferably used to perform clustering analysis on the first coating quality time series data X in the first sample subset. Specifically, the K-means clustering algorithm is used to cluster the first coating quality time series data X, and the DTW distance is used to measure the similarity between sequences. The DTW distance is the distance calculated by using the dynamic time warping algorithm (Dynamic Time Warping, DTW). The loss function of the model can be defined as the sum of squared errors SSE of the centers of the clusters to which each sequence belongs:

[0091]

[0092] In the formula, C i represents a cluster, k represents the number of cluster centers, p represents the sequence samples within the cluster, and o i represents the cluster centroid. DTW(p, o i ) represents the DTW distance from sequence p to the cluster center point.

[0093] Randomly select k samples as the centroids o i of the cluster C i (i = 1, 2,..., k), and perform sample clustering, that is, calculate the DTW distance from each sample to each centroid o i , and classify the sample into the cluster with the smallest DTW distance.

[0094] The following briefly explains the derivation process of the dynamic time warping algorithm. Assume that the sample sequence is a with a length of n, and the centroid sequence is b with a length of m. Then the goal of the dynamic time warping algorithm is to find an optimal path to minimize the cumulative distance between the two sequences. First, calculate the distance d(a_i, b_j) between each pair of points in the two sequences, and construct an n×m cumulative distance matrix D, where D(i, j) represents the minimum cumulative distance from (a_1, b_1) to (a_i, b_j). Then initialize the first row and the first column of the matrix D:

[0095]

[0096] Other elements are calculated using a recurrence formula:

[0097] D(i,j) = d(a_i,b_j)+min(D(i - 1,j),D(i,j - 1),D(i - 1,j - 1))

[0098] Backtrack from D(n,m) to D(1,1) to find the path that minimizes the cumulative distance, and thus the final DTW distance is D(n,m). For the specific reasoning process, refer to the dynamic time warping algorithm and it will not be elaborated here one by one.

[0099] Then perform centroid update, that is, calculate the current loss function, calculate the mean of each dimension of all samples within the cluster, and use this mean sequence as the new centroid of the cluster. The mean sequence can be expressed as:

[0100]

[0101] where CL i is the number of samples in cluster C i within.

[0102] Iteratively calculate the above sample clustering and centroid update until the loss function no longer changes.

[0103] Set k to different values and perform clustering respectively. Preferably, set k to 3, 5, 7, 9, 11, 13, 15, and use the above loss function to evaluate the clustering effect. For the clustering effects corresponding to each k value, make a line chart and use the elbow method to select the clustering centroid corresponding to the k value with the highest benefit. The selected clustering centroid is the clustering centroid corresponding to each sample type.

[0104] Based on the selected clustering centroids, classify the second sample subset, that is, classify the sample data in the second sample subset by calculating the similarity between each first coating quality time series data X in the second sample subset and each clustering centroid. In this embodiment, preferably, the dynamic time warping algorithm is used to calculate the DTW distance between the time series data and the clustering centroid as the similarity, and select the sample type corresponding to the centroid with the smallest DTW distance from the time series data as the type of this sample. The calculation process is the same as the above DTW distance calculation steps and will not be repeated here. It should be noted that the similarity calculation method in this embodiment is only a preference rather than a limitation, and other similarity calculation methods can also be used for classification, such as cosine similarity, Euclidean distance, Pearson correlation coefficient or Manhattan distance and other algorithms, and specific ones can be flexibly selected according to the actual situation.

[0105] Classify through the second subset of samples to obtain a subset of sample data corresponding to each data type. Of course, the classified samples in the first subset of samples can also be combined with the classified samples in the second subset of samples as the subset of sample data for this type of data.

[0106] In this embodiment, the time series prediction model for predicting the coating time series data is constructed using a neural network model such as a convolutional neural network model, a recurrent neural network model, or a long short-term memory neural network model, etc., and then each subset of sample data is used to train the time series prediction model. Here, a long short-term memory neural network model is preferably used to construct the time series prediction model. For each sample (X, Y) in each subset of sample data, the first coating quality time series data X is used as the input of the model, and the second coating quality time series data Y is used as the output of the model. Therefore, the loss function of the time series prediction model is defined as the weighted average error between the predicted result [mb′0…mb′ L-1 and the true value [mb0…mb L-1 , that is:

[0107]

[0108]

[0109] In the formula, n is the number of validation samples, e represents the error between the predicted sequence and the true sequence, and the closer the weighted average error LOSS is to zero, the better the prediction effect of the model.

[0110] In a preferred embodiment, the structures of the time series prediction models corresponding to each data type are the same. By training the time series prediction model with the same structure using different types of subsets of sample data, the time series prediction models corresponding to various types of data are obtained, and each time series prediction model is added to the time series prediction model set. Of course, each time series prediction model can also adopt different model structures, which are not overly restricted here, and the specific training steps of the model can refer to the conventional model training steps, which will not be elaborated one by one here.

[0111] After the above steps, the clustering centroids corresponding to each data type and the trained time series prediction model set are obtained. In actual application, first, the coating time series data of the battery electrode sheet is obtained, and the sample type of the coating time series data is determined by calculating the similarity between the coating time series data and the clustering centroids of various types. Here, the coating time series data refers to the coating quality data.

[0112] In a preferred embodiment, the present invention uses the dynamic time warping algorithm to calculate the DTW distance between the coating time series data and the centroid of each type of cluster, and uses the DTW distance as the similarity. The centroid of the cluster with the shortest DTW distance is used as the membership centroid of the coating time series data, and the sample type corresponding to the membership centroid is used as the sample type of the coating time series data. Then, based on the determined sample type, the time series prediction model corresponding to the sample type is selected from the set of time series prediction models, and the coating time series data is input into the time series prediction model to obtain the corresponding coating time series prediction data.

[0113] The coating time series prediction data is the predicted coating quality data. By analyzing this prediction data, the coating quality of the current battery electrode is determined, and it is judged whether it is necessary to adjust the production parameters of the coating device, so as to realize the guidance of the coating production. Preferably, according to the type of the selected coating quality data, a mapping relationship between the coating quality data and the production parameters of the coating device, such as pump speed and other parameters, can be established in advance. Through the coating time series prediction data and the corresponding mapping relationship, the rapid guidance of the production parameters of the coating device can be realized, thus solving the lag of the coating production quality adjustment and improving the quality and efficiency of the coating production.

[0114] For the problem that the prediction accuracy of the traditional method for predicting the coating quality of battery electrodes is not high and the detection is not timely, resulting in the lag of the coating production quality adjustment, the present invention effectively improves the accuracy of the prediction result of the coating quality data by establishing a high-quality sample data set and using a classified time series prediction model to predict the coating data, thus providing a timely and effective reference for the adjustment of the coating production parameters of the battery electrodes, solving the lag of the coating production quality adjustment, and further improving the quality and efficiency of the battery electrode coating production.

[0115] Please refer to Figure 2 , based on the same inventive concept, a prediction system for the coating quality of battery electrodes proposed in the second embodiment of the present invention includes:

[0116] A sample classification module 10, configured to obtain the coating time series data of the battery electrode, calculate the similarity between the coating time series data and the centroid of each preset sample type, and obtain the sample type of the coating time series data according to the similarity. The centroid of the cluster is obtained by performing a clustering analysis on the preset sample data set using a clustering algorithm, and the sample data in the sample data set includes coating quality data;

[0117] The timing prediction module 20 is configured to select a timing prediction model corresponding to the sample type from a pre-established set of timing prediction models, and input the coating timing data into the timing prediction model to obtain the coating timing prediction data. The set of timing prediction models includes several timing prediction models, and each timing prediction model corresponds to a sample type one by one. Each timing prediction model is constructed using a neural network model and trained using a sample data set;

[0118] The quality analysis module 30 is configured to determine the coating quality of the battery electrode based on the coating timing prediction data.

[0119] The technical features and technical effects of the prediction system for the coating quality of the battery electrode proposed in the embodiments of the present invention are the same as those of the method proposed in the embodiments of the present invention, and will not be elaborated here. Each module in the above prediction system for the coating quality of the battery electrode can be implemented in whole or in part by software, hardware, and their combinations. The above-mentioned modules can be embedded in the processor of the computer device in the form of hardware or be independent of it, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0120] In addition, an embodiment of the present invention also proposes a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0121] Please refer to Figure 3 , the internal structure diagram of the computer device in one embodiment. The computer device can specifically be a terminal or a server. The computer device includes a processor, a memory, a network interface, a display, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the prediction method for the coating quality of the battery electrode is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0122] Those of ordinary skill in the art can understand that Figure 3The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computing device may include more or fewer components than those shown in the figure, or combine some components, or have the same component arrangement.

[0123] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0124] In summary, an embodiment of the present invention provides a method, system, device, and storage medium for predicting the coating quality of battery electrodes. This method obtains the coating timing data of the battery electrodes, calculates the similarity between the coating timing data and the clustering centroids corresponding to each preset sample type, and obtains the sample type of the coating timing data according to the similarity. The clustering centroids are obtained by performing clustering analysis on a preset sample data set using a clustering algorithm. The sample data in the sample data set includes coating quality data; a timing prediction model corresponding to the sample type is selected from a pre-established set of timing prediction models, and the coating timing data is input into the timing prediction model to obtain coating timing prediction data. Among them, the set of timing prediction models includes several timing prediction models, and the timing prediction models correspond to the sample types one by one. Each timing prediction model is constructed using a neural network model and trained using the sample data set; according to the coating timing prediction data, the coating quality of the battery electrodes is determined. By establishing a high-quality sample data set and using a classified timing prediction model to predict the coating data, the present invention effectively improves the accuracy of the prediction result of the coating quality data, thereby providing a timely and effective reference for the adjustment of the coating production parameters of the battery electrodes. Through the method provided by the present invention, the lag in the adjustment of the coating production quality is effectively solved, and the quality and efficiency of the battery electrode coating production are improved.

[0125] Each embodiment in this specification is described in a progressive manner. For parts that are the same or similar in each embodiment, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and reference can be made to the corresponding part of the method embodiment for relevant content. It should be noted that the above technical features of the embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the above technical features in the embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0126] The above-described embodiments merely represent several preferred embodiments of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and substitutions can be made, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for predicting the coating quality of a battery electrode, characterized in that: include: Acquire coating time series data of a battery electrode sheet, calculate the similarity between the coating time series data and the cluster centroids corresponding to each preset sample type, and obtain the sample type of the coating time series data according to the similarity, wherein the cluster centroids are obtained by clustering a preset sample data set using a clustering algorithm, and the sample data in the sample data set includes coating quality data; Selecting a timing prediction model corresponding to the sample type from a pre-established timing prediction model set, and inputting the coating timing data into the timing prediction model to obtain coating timing prediction data, wherein the timing prediction model set includes a plurality of timing prediction models, the timing prediction models correspond to the sample types one by one, and each of the timing prediction models is constructed using a neural network model and trained using the sample data set; The coating quality of the battery electrode is determined based on the coating timing prediction data.

2. The method for predicting the coating quality of a battery electrode according to claim 1, characterized in that: The steps of constructing the sample data set include: Acquire historical coating data of the battery electrode, and preprocess the historical coating data to obtain a coating quality time series, wherein the historical coating data includes coating quality data, sampling time, coating speed, and coating die adjustment mark; The coating die head adjustment mark that is the same as the preset mark value is taken as the center point, and the coating quality time series is intercepted forward according to a preset length to obtain a plurality of coating quality time subsequences, wherein the preset length is determined based on the length of the coating oven; Divide the coating quality time subseries into equal lengths to obtain first coating quality time series data and second coating quality time series data, and establish a sample data set according to the first coating quality time series data and the second coating quality time series data; Dividing the sample data set into a first sample subset and a second sample subset according to a preset ratio; Using a K-means clustering algorithm based on dynamic time warping, cluster analysis is performed on each of the first coating quality time series data in the first sample subset to obtain a cluster centroid corresponding to each sample type; A dynamic time warping algorithm is used to calculate the first similarity between each of the first coating quality time series data and each of the cluster centroids in the second sample subset, and the second sample subset is classified according to the first similarity to obtain a sample data subset corresponding to each of the sample types.

3. The method for predicting the coating quality of a battery electrode according to claim 2, characterized in that: The step of preprocessing the historical coating data to obtain a coating quality time series comprises: Performing standardization processing on the historical coating data to obtain first standardized coating data; Calculating a sampling movement distance according to the sampling time and the coating speed, and updating the first standardized coating data according to the sampling movement distance to obtain second standardized coating data; The coating quality data in the second standardized coating data is updated according to a preset sampling frequency and the sampling moving distance, and a coating quality time series is established according to the updated coating quality data.

4. The method for predicting the coating quality of a battery electrode according to claim 3, characterized in that: The step of calculating the sampling movement distance according to the sampling time and the coating speed, and updating the first standardized coating data according to the sampling movement distance to obtain the second standardized coating data comprises: Calculating a sampling distance interval according to the sampling time and the coating speed, and establishing a sampling distance interval sequence according to the sampling time; The first sampling distance interval in the sampling distance interval sequence is set to zero, and the sampling distance interval sequence after being set to zero is accumulated and calculated to obtain a sampling movement distance sequence, wherein the sampling movement distance sequence is composed of the sampling movement distances corresponding to each sampling time; The sampling time and the coating speed in the first standardized coating data are replaced by the corresponding sampling moving distance to obtain second standardized coating data.

5. The method for predicting the coating quality of a battery electrode according to claim 3, characterized in that: The step of updating the coating quality data in the second standardized coating data according to the preset sampling frequency and the sampling moving distance includes: Taking the sampling movement distance as the dividend and the preset sampling frequency as the divisor, calculating the remainder between each sampling movement distance and the sampling frequency; Determine whether the remainder is equal to the remainder threshold; if not, perform linear interpolation update on the coating quality data corresponding to the sampling moving distance according to the sampling frequency and the sampling moving distance to obtain updated coating quality data; if yes, keep the coating quality data corresponding to the sampling moving distance unchanged; The updated coating quality data is expressed by the following formula: In the formula, m′ k represents the coating quality data updated for the kth time, f represents the sampling frequency, m k represents the kth coating quality data, m k-1 represents the k-1th coating quality data, l k represents the kth sample moving distance, l k-1 represents the k-1th sampling moving distance.

6. The method for predicting the coating quality of a battery electrode according to claim 2, characterized in that: The step of calculating the similarity between the coating time series data and the cluster centroids corresponding to each preset sample type, and obtaining the sample type of the coating time series data according to the similarity comprises: Using a dynamic time warping algorithm to calculate the second similarity between the coating time series data and each of the cluster centroids; According to the second similarity, the cluster centroid with the highest similarity to the coating time series data is selected from each of the cluster centroids as the belonging cluster centroid; The sample type of the coating time series data is determined according to the centroid of the belonging cluster.

7. The method for predicting the coating quality of a battery electrode according to claim 2, characterized in that: Each of the time series prediction models is trained using the corresponding sample data subset according to the sample type, wherein the first coating quality time series data in the sample data subset is sample data, and the second coating quality time series data in the sample data subset is label data.

8. A prediction system for battery pole piece coating quality, characterized in that: include: A sample classification module, used to obtain coating time series data of battery pole pieces, calculate the similarity between the coating time series data and the cluster centroids corresponding to each preset sample type, and obtain the sample type of the coating time series data according to the similarity, wherein the cluster centroids are obtained by clustering a preset sample data set using a clustering algorithm, and the sample data in the sample data set includes coating quality data; A timing prediction module, used for selecting a timing prediction model corresponding to the sample type from a pre-established timing prediction model set, and inputting the coating timing data into the timing prediction model to obtain coating timing prediction data, wherein the timing prediction model set includes a plurality of timing prediction models, the timing prediction models correspond to the sample types one by one, and each of the timing prediction models is constructed using a neural network model and trained using the sample data set; The quality analysis module is used to determine the coating quality of the battery electrode according to the coating timing prediction data.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images