Method, device and computer device for constructing capacity retention rate estimation model

Abstract

The application relates to a capacity retention rate estimation model construction method, device and computer equipment. The method comprises the following steps: obtaining a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each feature data set comprises a plurality of sample features; training an initial random forest model according to the plurality of feature data sets to construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features. The feature weights are introduced into the loss of node splitting, the sensitivity of the model to important features is improved, the efficiency of training the initial random forest model is improved, and the accuracy of the target prediction model is improved; and the initial random forest model is trained according to the plurality of feature data sets, the nonlinear relationship between the sample features is fully utilized, and the prediction accuracy of the target prediction model for the later capacity retention rate of the battery is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium ion batteries, in particular to a capacity retention rate estimation model construction method, device and computer equipment. BACKGROUND

[0002] Lithium ion batteries have the characteristics of high energy density, long cycle life and environmental friendliness, and are widely used in portable electronic devices, electric vehicles and energy storage systems, etc. However, as the use time of lithium ion batteries increases, the capacity of lithium ion batteries will gradually decrease, resulting in the termination of the use of the battery. Therefore, estimating the capacity retention rate of lithium ion batteries is of great significance for evaluating battery performance, guiding battery design, optimizing battery management, discovering potential problems and reducing development cost and time.

[0003] Generally, the capacity retention rate of lithium ion batteries is estimated by collecting data throughout the life cycle of lithium ion batteries.

[0004] However, the conventional technology has the problem of low efficiency in estimating the capacity retention rate of lithium ion batteries. SUMMARY

[0005] Therefore, it is necessary to provide a capacity retention rate estimation model construction method, device and computer equipment capable of improving the estimation efficiency of the capacity retention rate of the battery in view of the above technical problems.

[0006] In a first aspect, the present application provides a capacity retention rate estimation model construction method, comprising:

[0007] According to the charge-discharge data of a plurality of sample batteries, a plurality of feature data sets are obtained; each of the feature data sets comprises a plurality of sample features;

[0008] According to the plurality of feature data sets, an initial random forest model is trained to construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of each sample feature.

[0009] In one embodiment, the training of the initial random forest model according to the plurality of feature data sets to construct the target prediction model for predicting the capacity retention rate of any target battery comprises:

[0010] According to the plurality of feature data sets, an initial random forest model is trained to obtain a prediction model corresponding to each feature data set;

[0011] According to the prediction model corresponding to each feature data set and the model weights of each prediction model, a target prediction model for predicting the capacity retention rate of any target battery is constructed.

[0012] In one of the embodiments, the training of the initial random forest model according to the plurality of feature data sets to obtain a prediction model corresponding to each of the feature data sets comprises:

[0013] inputting the sample features into the initial random forest model to obtain an initial prediction result; the initial random forest model is composed of a plurality of leaf nodes and root nodes;

[0014] determining a node loss of each of the leaf nodes in the initial random forest model according to the feature weight of each of the sample features and the initial prediction result;

[0015] training the initial random forest model according to each of the node losses to obtain a prediction model corresponding to each of the feature data sets.

[0016] In one of the embodiments, the determining of the node loss of each of the leaf nodes in the initial random forest model according to the feature weight of each of the sample features and the initial prediction result comprises:

[0017] determining a first weighted mean square error of each of the leaf nodes in the initial random forest model and a second weighted mean square error of a child node associated with each of the leaf nodes according to the feature weight of each of the sample features;

[0018] determining the node loss of each of the leaf nodes in the initial random forest model according to the first weighted mean square error of each of the leaf nodes and the second weighted mean square error of each of the child nodes.

[0019] In one of the embodiments, the method further comprises:

[0020] determining the correlation between each of the sample features and the capacity retention rate;

[0021] determining the sample weight of each of the sample features according to the correlation between each of the sample features and the capacity retention rate.

[0022] In one of the embodiments, the method further comprises:

[0023] obtaining the charge-discharge data of the target battery, and determining a feature data set corresponding to the target battery according to the charge-discharge data;

[0024] inputting the feature data set corresponding to the target battery into the target prediction model to output the capacity retention rate of the target battery.

[0025] In a second aspect, the application further provides a construction device of a capacity retention rate estimation model, comprising:

[0026] The first obtaining module is configured to obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each of the feature data sets comprises a plurality of sample features;

[0027] The constructing module is configured to train an initial random forest model according to the plurality of feature data sets, and construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0028] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0029] The first obtaining module is configured to obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each of the feature data sets comprises a plurality of sample features;

[0030] The constructing module is configured to train an initial random forest model according to the plurality of feature data sets, and construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0031] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0032] The first obtaining module is configured to obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each of the feature data sets comprises a plurality of sample features;

[0033] The constructing module is configured to train an initial random forest model according to the plurality of feature data sets, and construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0034] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the following steps:

[0035] The first obtaining module is configured to obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each of the feature data sets comprises a plurality of sample features;

[0036] The constructing module is configured to train an initial random forest model according to the plurality of feature data sets, and construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0037] The method, device and computer equipment for constructing the capacity retention rate estimation model, obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each feature data set includes a plurality of sample features; an initial random forest model is trained according to the plurality of feature data sets, and a target prediction model for predicting the capacity retention rate of any target battery is constructed; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features. The feature weights are introduced into the loss of node splitting, the sensitivity of the model to important features is improved, the efficiency of training the initial random forest model is improved, and the accuracy of the target prediction model is improved; and the initial random forest model is trained according to the plurality of feature data sets, the nonlinear relationship between the sample features is fully utilized, and the prediction accuracy of the target prediction model for the later capacity retention rate of the battery is improved. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without creative labor.

[0039] Figure 1 The application environment diagram of the capacity retention rate estimation model construction method in one embodiment;

[0040] Figure 2 The flowchart of the capacity retention rate estimation model construction method in one embodiment;

[0041] Figure 3 The flowchart of the capacity retention rate estimation model construction method in another embodiment;

[0042] Figure 4 The performance diagram of each prediction model for predicting according to the training set in one embodiment;

[0043] Figure 5 The performance diagram of each prediction model for predicting according to the test set in one embodiment;

[0044] Figure 6 The flowchart of the capacity retention rate estimation model construction method in another embodiment;

[0045] Figure 7 The flowchart of the capacity retention rate estimation model construction method in another embodiment;

[0046] Figure 8 The flowchart of the capacity retention rate estimation model construction method in another embodiment;

[0047] Figure 9 a flowchart of a method for constructing a capacity retention rate estimation model in another embodiment;

[0048] Figure 10 a flowchart of a method for constructing a capacity retention rate estimation model in another embodiment;

[0049] Figure 11 a structural block diagram of a device for constructing a capacity retention rate estimation model in an embodiment. DETAILED DESCRIPTION

[0050] In order to make the purposes, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0051] The method for constructing a capacity retention rate estimation model provided by the embodiments of the present application can be applied in an application environment as shown in Figure 1 The computer device can be a server, and its internal structure diagram can be as shown in Figure 1 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store construction data of a capacity retention rate estimation model. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a method for constructing a capacity retention rate estimation model.

[0052] Those skilled in the art can understand that Figure 1 The structure shown in the above is only a block diagram of part of the structure related to the scheme of the present application, and does not limit the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0053] In an embodiment, as Figure 2As shown, a capacity retention estimation model construction method is provided, and the method is applied to Figure 1 a server in

[0054] S201, obtaining a plurality of feature data sets according to the charge and discharge data of a plurality of sample batteries; each feature data set includes a plurality of sample features.

[0055] The sample battery can be a lithium ion battery, and the charge and discharge data of the sample battery can be long-term electrochemical performance data of the battery, for example, the charge and discharge data of the sample battery can be data of 50-500 complete charge and discharge cycles; the plurality of sample features can include the maximum value, the minimum value, the difference value, the average value, the variance, and the change trend of each index with the increase of the charge and discharge cycle number, the capacity decay trend and the variance, the extreme point of the differential capacity test curve, the change trend, and the n-order interval, etc.

[0056] In the embodiments of the present application, the server can obtain the charge and discharge data of a plurality of sample batteries in the data acquisition device of the sample battery in advance and store them in the local database; or the server can obtain the charge and discharge data of a plurality of sample batteries from the data storage device. Further, the charge and discharge data of the plurality of sample batteries is cleaned to remove irrelevant or redundant features in the charge and discharge data of the plurality of sample batteries, and the missing values and abnormal values are processed to obtain the cleaned charge and discharge data of each sample battery, so that the sample data of each sample battery is extracted from the cleaned charge and discharge data of each sample battery, the sample data of each battery includes a plurality of sample features, and the sample data of all sample batteries is sampled to generate a plurality of feature data sets. Optionally, the sample data of all batteries can be represented as wherein is the sample data of the i-th sample battery, is the target variable of the i-th sample data, i.e., the capacity retention rate, n sample data are randomly extracted with replacement in multiple D, and a plurality of feature data sets are obtained, i.e., each feature data set includes n sample data.

[0057] Exemplarily, irrelevant or redundant features in the charge and discharge data of the plurality of sample batteries can be deleted, such as battery number and features directly related to later capacity retention rate, to avoid information leakage and model overfitting. The missing values and abnormal values can be processed by first checking whether there are missing values or abnormal values in the data. For missing values, the deletion sample or interpolation method can be used for processing; for abnormal values, statistical methods can be used for detection and processing, such as using box plot or Z-score method for detection and processing.

[0058] Optionally, feature standardization can be performed on multiple sample features in multiple feature datasets to eliminate the dimensional differences between different features, resulting in multiple processed feature datasets. These multiple feature datasets can then be used to train the initial random forest model. For example, a standardization method can be used to convert each feature into data with a mean of 0 and a standard deviation of 1. For instance, standardization methods can be shown in Equations 1 to 3:

[0059] (Equation 1)

[0060] (Equation 2)

[0061] (Equation 3)

[0062] in, For the i-th feature in the dataset, the j-th feature in the dataset for The corresponding standardized data, It is the mean of the j-th feature in all feature datasets. is the standard deviation of the j-th feature, and n is the data in the feature dataset, i.e., the number of sample batteries.

[0063] Optionally, feature extraction can be performed on the charge-discharge data of multiple sample batteries to obtain multiple candidate sample features. The correlation between each candidate sample feature and the capacity retention rate can be determined, thereby removing candidate sample features with a correlation no greater than a correlation threshold and retaining those with a correlation greater than the threshold, resulting in multiple sample features and forming multiple feature datasets. For example, the correlation between each candidate sample feature and the capacity retention rate can be determined based on the Pearson correlation coefficient. Using data processing methods such as standardization and correlation analysis can reduce the data complexity of the model.

[0064] S202, the initial random forest model is trained based on multiple feature datasets to construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model during training is determined based on the feature weights of each sample feature.

[0065] In this embodiment of the application, at each node of the initial random forest model, a feature subset is determined using a weighted probability sampling method. Specifically, at each node of the initial random forest model, the feature subset is determined according to the feature weights. The weight distribution is constructed, and the sample features in the feature data set are weighted and randomly sampled, m sample features are selected to form a feature subset S, further, the loss of node splitting is determined according to the feature subset of the current node, the feature weight is introduced into the loss of node splitting, and the sensitivity of the model to important features is improved. Optionally, the node splitting direction of each node can be determined according to the loss corresponding to the leaf node of each node. In the embodiment of the application, the node splitting is recursively performed until the stopping condition of node splitting is met, and the target prediction model is obtained. Optionally, the stopping condition can be that the random forest model reaches the maximum depth, or the number of leaf node samples is less than the minimum sample number. Optionally, the feature subset can be determined by weighted probability sampling on multiple feature data sets; or the feature subset can be determined by weighted probability sampling on any feature data set.

[0066] In the method for constructing the capacity retention rate estimation model, a plurality of feature data sets are obtained according to the charge and discharge data of a plurality of sample batteries; each feature data set includes a plurality of sample features; an initial random forest model is trained according to the plurality of feature data sets, and a target prediction model for predicting the capacity retention rate of any target battery is constructed; and the loss of the initial random forest model in the training process is determined according to the feature weight of each sample feature. The feature weight is introduced into the loss of node splitting, the sensitivity of the model to important features is improved, the efficiency of training the initial random forest model is improved, and the accuracy of the target prediction model is improved; and the initial random forest model is trained according to the plurality of feature data sets, the nonlinear relationship between the sample features is fully utilized, and the prediction accuracy of the target prediction model for the later capacity retention rate of the battery is improved.

[0067] In one embodiment, an implementation of S202 is provided as shown in Figure 3 According to the plurality of feature data sets, the initial random forest model is trained, and the target prediction model for predicting the capacity retention rate of any target battery is constructed.

[0068] S301, according to the plurality of feature data sets, the initial random forest model is trained, and the target prediction model for predicting the capacity retention rate of any target battery is constructed.

[0069] In one embodiment, for each feature data set, a weighted probability sampling method is used to determine a feature subset at each node of the initial random forest model, further, the loss of node splitting is determined according to the feature subset of the current node, the loss of each node is determined to determine the splitting direction of the node, until the stopping condition of node splitting is met, and the prediction model corresponding to each feature data set is obtained.

[0070] S302, constructing a target prediction model for predicting the capacity retention rate of any target battery according to the prediction model corresponding to each feature data set and the model weight of each prediction model.

[0071] In the embodiments of the present application, if the number of prediction models is T, the target prediction model can be as shown in formula 4:

[0072] (Formula 4)

[0073] wherein, is the tth prediction model, is the model weight of the tth prediction model.

[0074] As an optional implementation, the weights of each feature data set can be pre-configured, so as to determine the model weight of each prediction model according to the weight of each feature data set, for example, the weight of the feature data set is determined as the model weight of the corresponding prediction model.

[0075] As another optional implementation, part of the plurality of feature data sets can be used as a training set to train a plurality of prediction models, and the remaining feature data sets in the plurality of feature data sets can be used as a test set to test the performance of the plurality of prediction models, and the model weight of each prediction model can be determined according to the performance test result of each prediction model, for example, the prediction accuracy of each prediction model in the performance test is used as the model weight of each prediction model. Optionally, the performance of each prediction model in predicting the capacity retention rate of the 100th charge and discharge cycle period in the training set is as shown in Figure 4 , the root mean square error (Root Mean Square Error, RMSE) is 0.003403, the mean absolute percentage error (Mean Absolute Percentage Error, MAPE) is 0.25%, and the coefficient of determination R 2 is 0.8779; the performance of the 100th charge and discharge cycle period capacity retention rate prediction model of each prediction model in the prediction set is as shown in Figure 5 , the RMSE is 0.004519, the MAPE is 0.46%, and the R 2The value is 0.8003. For example, at the 100th week of the charge-discharge cycle for each sample, the actual value of the capacity retention rate of sample battery 0 is 0.941, and the predicted value of the target prediction model for the capacity retention rate of sample battery 0 is 0.933; the actual value of the capacity retention rate of sample battery 1 is 0.917, and the predicted value of the target prediction model for the capacity retention rate of sample battery 1 is 0.928; the actual value of the capacity retention rate of sample battery 2 is 0.938, and the predicted value of the target prediction model for the capacity retention rate of sample battery 2 is 0.947; the actual value of the capacity retention rate of sample battery 3 is 0.953, and the predicted value of the target prediction model for the capacity retention rate of sample battery 3 is 0.8003. The predicted value for the measured capacity retention rate was 0.947; the actual value for the capacity retention rate of sample battery 4 was 0.935, and the predicted value for the capacity retention rate of sample battery 4 by the target prediction model was 0.931; the actual value for the capacity retention rate of sample battery 5 was 0.949, and the predicted value for the capacity retention rate of sample battery 5 by the target prediction model was 0.944; the actual value for the capacity retention rate of sample battery 6 was 0.926, and the predicted value for the capacity retention rate of sample battery 6 by the target prediction model was 0.93; the actual value for the capacity retention rate of sample battery 7 was 0.943, and the predicted value for the capacity retention rate of sample battery 7 by the target prediction model was 0.941.

[0076] In the above application embodiments, multiple prediction models are trained. Compared with training a single prediction model, the target prediction model obtained by multiple prediction models has higher accuracy. Furthermore, by combining the model weights of each prediction model, the prediction accuracy of the target prediction model is further improved.

[0077] In one embodiment, one implementation of the above-described S301 is provided, such as... Figure 6 As shown, the above-mentioned "training the initial random forest model based on multiple feature datasets to obtain the prediction model corresponding to each feature dataset" includes:

[0078] S401, input the sample features into the initial random forest model to obtain the initial prediction results; the initial random forest model consists of multiple leaf nodes and a root node.

[0079] In this embodiment of the application, for each set of feature datasets, the sample data in the feature datasets are input into the random forest model, so that the initial random forest model splits the root node into leaf nodes according to the sample features and predicts the initial prediction result.

[0080] S402, based on the feature weights of each sample feature and the initial prediction results, determine the node loss of each leaf node in the initial random forest model.

[0081] In the embodiments of the present application, the weighted minimum mean square error of each leaf node and the weighted minimum mean square error of the child node of each leaf node performing node splitting can be determined according to the feature weight of each sample feature, so as to determine the node loss of the leaf node according to the weighted minimum mean square error of the leaf node and the weighted minimum mean square error of the child node.

[0082] Optionally, as shown in Figure 7 The determination process of the feature weight of each sample feature can include:

[0083] S404, determining the correlation between each sample feature and the capacity retention rate.

[0084] S405, determining the sample weight of each sample feature according to the correlation between each sample feature and the capacity retention rate.

[0085] In the embodiments of the present application, the correlation between each sample feature and the capacity retention rate can be determined according to the Pearson correlation coefficient between each sample feature and the capacity retention rate. Further, the sum of all correlations can be determined first to obtain a correlation sum, and then the ratio between the correlation between each sample feature and the capacity retention rate and the correlation sum is determined as the sample weight corresponding to each sample feature.

[0086] S403, training the initial random forest model according to each node loss to obtain a prediction model corresponding to each feature data set.

[0087] In the embodiments of the present application, for the splitting of the root node and each leaf node, the sample feature and the splitting point corresponding to the maximum node loss are selected for node splitting to determine the split leaf node, and the step of "inputting the sample feature into the initial random forest model to obtain the initial prediction result" is returned to be executed until the stop condition of node splitting is met, thereby obtaining a prediction model corresponding to each feature data set.

[0088] In the above application embodiments, the feature weight is introduced into the loss of node splitting, the sensitivity of the model to important features is improved, thereby improving the efficiency of training the initial random forest model and improving the accuracy of the target prediction model.

[0089] In one embodiment, an implementation of S403 is provided, as shown in Figure 8 The "determining the node loss of each leaf node in the initial random forest model according to the feature weight of each sample feature and the initial prediction result" includes:

[0090] S501, determining the first weighted mean square error of each leaf node in the initial random forest model and the second weighted mean square error of the child node associated with each leaf node according to the feature weight of each sample feature.

[0091] In this embodiment, a first weighted mean square error for each leaf node is determined based on the feature weights of each sample feature, and a second weighted mean square error for the child nodes associated with each leaf node is determined based on the feature weights of the sample features. The methods for determining the first and second weighted mean square errors are the same. For example, the first weighted mean square error for each leaf node can be as shown in Equation 5:

[0092] (Equation 5)

[0093] in, Let N be the number of samples at node N. For the predicted value of sample data i, The sample weights are the features of the j-th sample.

[0094] S502, based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node, determine the node loss of each leaf node in the initial random forest model.

[0095] In this embodiment, the node loss of each leaf node can be expressed as shown in Equation 6:

[0096] (Equation 6)

[0097] in, Let N be the first weighted mean square error of node N. Let N be the left child node. Let N be the right child node. For nodes The number of samples, For nodes The number of samples, child node The second weighted mean square error, child node The second weighted mean square error.

[0098] In the above application embodiments, each leaf node is determined based on the feature weights of the sample features, and the weighted mean square error of the child nodes associated with each leaf node is determined, so that the node splitting of the random forest model is more closely matched with the target battery being tested, thereby improving the accuracy of the target prediction model in predicting the battery's capacity retention rate.

[0099] In one embodiment, such as Figure 9 As shown, the method for constructing the above-mentioned capacity retention rate estimation model also includes:

[0100] S203, acquire the charge and discharge data of the target battery, and determine the feature dataset corresponding to the target battery based on the charge and discharge data.

[0101] The target battery can be a lithium ion battery, and the charge-discharge data of the target battery can be early electrochemical performance data of the battery, for example, the charge-discharge data of the target battery can be data of complete charge-discharge cycles in the first 50 weeks. The feature data set can include maximum, minimum, difference, average, variance of charge-discharge voltage, and change trend of each index with the increase of the number of charge-discharge cycles, capacity decay trend and variance, extreme point of differential capacity test curve, change trend and n-order interval, etc.

[0102] In the embodiments of the present application, the charge-discharge data of the target battery is cleaned, irrelevant or redundant features in the charge-discharge data of the target battery are removed, and missing values and abnormal values are processed to obtain cleaned charge-discharge data. Then, the cleaned charge-discharge data is subjected to feature extraction to obtain a feature data set.

[0103] Optionally, the charge-discharge data of the target battery is subjected to feature extraction to obtain a plurality of candidate features. According to the correlation between each sample feature of the sample battery and the capacity retention rate, the candidate features of the target battery are screened to obtain screened candidate features, forming a feature data set.

[0104] S204, inputting the feature data set corresponding to the target battery into the target prediction model to output the capacity retention rate of the target battery.

[0105] In the embodiments of the present application, the target prediction model includes a plurality of trained random forest models. The feature data set is input into the plurality of trained random forest models in the target prediction model for prediction to obtain the prediction results of each random forest model. Then, the prediction results are weighted and averaged to obtain the capacity retention rate of the target battery.

[0106] In the embodiments of the present application, the early electrochemical performance data of the battery can be used to predict the capacity retention rate in the later stage, without the need for long-term cycle test, thereby shortening the time of battery performance test.

[0107] In one embodiment, a method for constructing a complete capacity retention rate estimation model is provided, as shown in Figure 10 The above method includes:

[0108] S1, obtaining a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each feature data set includes a plurality of sample features.

[0109] S2, determining the correlation between each sample feature and the capacity retention rate.

[0110] S3, determining the sample weight of each sample feature according to the correlation between each sample feature and the capacity retention rate.

[0111] S4. Input the sample features into the initial random forest model to obtain the initial prediction results; the initial random forest model consists of multiple leaf nodes and a root node.

[0112] S5. Based on the feature weights of each sample feature, determine the first weighted mean square error of each leaf node in the initial random forest model, and the second weighted mean square error of the child nodes associated with each leaf node.

[0113] S6. Based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node, determine the node loss of each leaf node in the initial random forest model.

[0114] S7. Train the initial random forest model based on the loss of each node to obtain the prediction model corresponding to each feature dataset.

[0115] S8. Based on the prediction models corresponding to each feature dataset and the model weights of each prediction model, a target prediction model for predicting the capacity retention rate of any target battery is constructed.

[0116] S9, acquire the charge and discharge data of the target battery, and determine the feature dataset corresponding to the target battery based on the charge and discharge data.

[0117] S10: Input the feature dataset corresponding to the target battery into the target prediction model and output the capacity retention rate of the target battery.

[0118] In the above method for constructing the capacity retention estimation model, multiple feature datasets are obtained based on the charge and discharge data of multiple sample batteries. Each feature dataset includes multiple sample features. An initial random forest model is trained based on these multiple feature datasets to construct a target prediction model for predicting the capacity retention of any target battery. The loss of the initial random forest model during training is determined based on the feature weights of each sample feature. Introducing feature weights into the loss of node splitting improves the model's sensitivity to important features, thereby improving the efficiency of training the initial random forest model and the accuracy of the target prediction model. Furthermore, training the initial random forest model based on multiple feature datasets fully utilizes the nonlinear relationships between sample features, improving the prediction accuracy of the target prediction model for the battery's later capacity retention.

[0119] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not strictly limited in sequence, and the steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time but can be executed at different times, and the execution of the steps or stages is not necessarily sequential but can be performed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0120] Based on the same inventive concept, the embodiments of the present application also provide a capacity retention rate estimation model construction device for implementing the capacity retention rate estimation model construction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more capacity retention rate estimation model construction device embodiments provided below can refer to the limitations of the capacity retention rate estimation model construction method described above, which will not be repeated here.

[0121] In one embodiment, as shown in Figure 11 A capacity retention rate estimation model construction device is provided, comprising: a first acquisition module 10 and a construction module 11, wherein:

[0122] The first acquisition module 10 is configured to acquire a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each feature data set includes a plurality of sample features.

[0123] The construction module 11 is configured to train an initial random forest model according to the plurality of feature data sets, and construct a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0124] In one embodiment, the construction module 11 described above comprises a training unit and a construction unit, wherein:

[0125] The training unit is configured to train an initial random forest model according to a plurality of feature data sets to obtain a prediction model corresponding to each feature data set.

[0126] The construction unit is configured to construct a target prediction model for predicting the capacity retention rate of any target battery according to the prediction model corresponding to each feature data set and the model weights of each prediction model.

[0127] In an embodiment, the training unit is specifically configured to input the sample features into the initial random forest model to obtain initial prediction results; the initial random forest model is composed of multiple leaf nodes and root nodes; node losses of the leaf nodes in the initial random forest model are determined according to the feature weights of the sample features and the initial prediction results; the initial random forest model is trained according to the node losses to obtain the prediction model corresponding to each feature dataset.

[0128] In an embodiment, the training unit is specifically configured to determine first weighted mean square errors of the leaf nodes in the initial random forest model and second weighted mean square errors of the child nodes associated with the leaf nodes according to the feature weights of the sample features; and determine the node losses of the leaf nodes in the initial random forest model according to the first weighted mean square errors of the leaf nodes and the second weighted mean square errors of the child nodes.

[0129] In an embodiment, the training unit is specifically configured to determine the correlations between the sample features and the capacity retention rates; and determine the sample weights of the sample features according to the correlations between the sample features and the capacity retention rates.

[0130] In an embodiment, the construction apparatus of the capacity retention rate estimation model further includes a second acquisition module and an output module, wherein:

[0131] The second acquisition module is configured to acquire the charge-discharge data of the target battery, and determine the feature dataset corresponding to the target battery according to the charge-discharge data.

[0132] The output module is configured to input the feature dataset corresponding to the target battery into the target prediction model, and output the capacity retention rate of the target battery.

[0133] The modules in the construction apparatus of the capacity retention rate estimation model can be all or partially realized by software, hardware, and combinations thereof. The modules can be embedded in or independent of a processor in a computer device in a hardware form, or stored in a memory in a computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the modules.

[0134] In an embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0135] A plurality of feature datasets are acquired according to charge-discharge data of a plurality of sample batteries; each feature dataset includes a plurality of sample features;

[0136] The initial random forest model is trained according to a plurality of feature data sets, and a target prediction model for predicting the capacity retention rate of any target battery is constructed; and the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0137] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0138] The initial random forest model is trained according to a plurality of feature data sets, and a target prediction model for predicting the capacity retention rate of any target battery is constructed; and the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0139] The initial random forest model is trained according to a plurality of feature data sets, and a target prediction model for predicting the capacity retention rate of any target battery is constructed; and the loss of the initial random forest model in the training process is determined according to the feature weights of the sample features.

[0140] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0141] The sample features are input into the initial random forest model to obtain an initial prediction result; the initial random forest model is composed of a plurality of leaf nodes and root nodes;

[0142] The node loss of each leaf node in the initial random forest model is determined according to the feature weights of the sample features and the initial prediction result;

[0143] The initial random forest model is trained according to the node loss to obtain the prediction model corresponding to each feature data set.

[0144] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0145] The first weighted mean square error of each leaf node in the initial random forest model and the second weighted mean square error of the child nodes associated with each leaf node are determined according to the feature weights of the sample features;

[0146] The node loss of each leaf node in the initial random forest model is determined according to the first weighted mean square error of each leaf node and the second weighted mean square error of each child node.

[0147] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0148] The correlation of each sample feature with the capacity retention rate is determined;

[0149] The sample weight of each sample feature is determined according to the correlation of each sample feature with the capacity retention rate.

[0150] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0151] obtaining charging and discharging data of a target battery, and determining a feature data set corresponding to the target battery according to the charging and discharging data;

[0152] inputting the feature data set corresponding to the target battery into a target prediction model, and outputting a capacity retention rate of the target battery.

[0153] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the following steps:

[0154] obtaining a plurality of feature data sets according to charging and discharging data of a plurality of sample batteries; each feature data set includes a plurality of sample features;

[0155] training an initial random forest model according to the plurality of feature data sets, and constructing a target prediction model for predicting a capacity retention rate of any target battery; a loss of the initial random forest model in a training process is determined according to feature weights of the sample features.

[0156] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0157] training the initial random forest model according to the plurality of feature data sets, and obtaining a prediction model corresponding to each feature data set;

[0158] constructing the target prediction model for predicting the capacity retention rate of any target battery according to the prediction model corresponding to each feature data set and model weights of the prediction models.

[0159] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0160] inputting the sample features into the initial random forest model to obtain an initial prediction result; the initial random forest model is composed of a plurality of leaf nodes and a root node;

[0161] determining node losses of the leaf nodes in the initial random forest model according to the feature weights of the sample features and the initial prediction result;

[0162] training the initial random forest model according to the node losses to obtain the prediction model corresponding to each feature data set.

[0163] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0164] determining a first weighted mean square error of each leaf node in the initial random forest model and a second weighted mean square error of a child node associated with each leaf node according to the feature weights of the sample features;

[0165] Determine the node loss of each leaf node in the initial random forest model according to the first weighted mean square error of each leaf node and the second weighted mean square error of each child node.

[0166] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0167] Determine the correlation of each sample feature with the capacity retention rate;

[0168] Determine the sample weight of each sample feature according to the correlation of each sample feature with the capacity retention rate.

[0169] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0170] Obtain the charge-discharge data of the target battery, and determine the feature data set corresponding to the target battery according to the charge-discharge data;

[0171] Input the feature data set corresponding to the target battery into the target prediction model, and output to obtain the capacity retention rate of the target battery.

[0172] In one embodiment, a computer program product is provided, comprising a computer program which, when executed by the processor, implements the following steps:

[0173] Obtain a plurality of feature data sets according to the charge-discharge data of a plurality of sample batteries; each feature data set comprises a plurality of sample features;

[0174] Train the initial random forest model according to the plurality of feature data sets, and build to obtain a target prediction model for predicting the capacity retention rate of any target battery; the loss of the initial random forest model in the training process is determined according to the feature weight of each sample feature.

[0175] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0176] Train the initial random forest model according to the plurality of feature data sets to obtain a prediction model corresponding to each feature data set;

[0177] According to the prediction model corresponding to each feature data set and the model weight of each prediction model, a target prediction model for predicting the capacity retention rate of any target battery is built.

[0178] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0179] Input the sample feature into the initial random forest model to obtain an initial prediction result; the initial random forest model is composed of a plurality of leaf nodes and a root node;

[0180] Based on the feature weights of each sample feature and the initial prediction results, determine the node loss of each leaf node in the initial random forest model;

[0181] The initial random forest model is trained based on the loss of each node to obtain the prediction model corresponding to each feature dataset.

[0182] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0183] Based on the feature weights of each sample feature, the first weighted mean square error of each leaf node and the second weighted mean square error of the child nodes associated with each leaf node in the initial random forest model are determined.

[0184] The node loss of each leaf node in the initial random forest model is determined based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node.

[0185] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0186] Determine the correlation between the characteristics of each sample and the capacity retention rate;

[0187] The sample weights of each sample feature are determined based on the correlation between each sample feature and the capacity retention rate.

[0188] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0189] Acquire the charge and discharge data of the target battery, and determine the corresponding feature dataset of the target battery based on the charge and discharge data;

[0190] The feature dataset corresponding to the target battery is input into the target prediction model, and the capacity retention rate of the target battery is output.

[0191] It should be noted that the data involved in this application (including but not limited to data used for analysis, data stored, data displayed, etc.) are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0192] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0193] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0194] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for constructing a capacity retention rate estimation model, characterized in that, The method includes: Multiple feature datasets are obtained based on the charge and discharge data of multiple sample batteries; each feature dataset includes multiple sample features; The initial random forest model is trained based on the multiple feature datasets to construct a target prediction model for predicting the capacity retention rate of any target battery. At each node of the initial random forest model, a subset of features is determined from the feature dataset using a weighted probability sampling method. The node loss of the leaf node of the node is determined based on the feature subset. The node splitting direction is determined based on the node loss corresponding to each leaf node. The process of determining the node loss of each leaf node includes: determining the first weighted mean square error of each leaf node and the second weighted mean square error of the child nodes associated with each leaf node in the initial random forest model based on the feature weights of each sample feature and the initial prediction result; determining the node loss of each leaf node in the initial random forest model based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node; the initial prediction result is obtained based on the sample features in the feature subset; The step of determining the node loss of each leaf node in the initial random forest model based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node includes: in, The first weighted mean square error of the leaf node. It is the left child node of the leaf node. This is the right child node of the leaf node. For the The number of samples, For the The number of samples, For the The second weighted mean square error, For the The method for determining the second weighted mean square error of each leaf node and the second weighted mean square error of the child nodes associated with each leaf node includes: in, Let N be the number of samples at node N. For the predicted value of sample data i, The sample weights are the features of the j-th sample.

2. The method according to claim 1, characterized in that, The step of training the initial random forest model based on the multiple feature datasets to construct a target prediction model for predicting the capacity retention rate of any target battery includes: The initial random forest model is trained based on the multiple feature datasets to obtain the prediction model corresponding to each feature dataset; Based on the prediction models corresponding to each of the feature datasets and the model weights of each prediction model, a target prediction model for predicting the capacity retention rate of any target battery is constructed.

3. The method according to claim 2, characterized in that, The step of training an initial random forest model based on the multiple feature datasets to obtain a prediction model corresponding to each feature dataset includes: The sample features are input into the initial random forest model to obtain initial prediction results; the initial random forest model consists of multiple leaf nodes and a root node. Based on the feature weights of each sample feature and the initial prediction results, determine the node loss of each leaf node in the initial random forest model; The initial random forest model is trained based on the loss of each node to obtain the prediction model corresponding to each feature dataset.

4. The method according to claim 3, characterized in that, The method further includes: Determine the correlation between each of the sample characteristics and the capacity retention rate; The sample weights of each sample feature are determined based on their correlation with the capacity retention rate.

5. The method according to any one of claims 1-3, characterized in that, The method further includes: Acquire the charge and discharge data of the target battery, and determine the feature dataset corresponding to the target battery based on the charge and discharge data; The feature dataset corresponding to the target battery is input into the target prediction model, and the capacity retention rate of the target battery is output.

6. An apparatus for constructing a capacity retention rate estimation model, characterized in that, The device includes: The first acquisition module is used to acquire multiple feature datasets based on the charge and discharge data of multiple sample batteries; each feature dataset includes multiple sample features; The construction module is used to train the initial random forest model based on the multiple feature datasets to construct a target prediction model for predicting the capacity retention rate of any target battery. At each node of the initial random forest model, a feature subset is determined from the feature dataset using a weighted probability sampling method. The node loss of the leaf node of the node is determined based on the feature subset. The node splitting direction is determined based on the node loss corresponding to each leaf node. The process of determining the node loss of each leaf node in the initial random forest model during training includes: determining the first weighted mean square error of each leaf node and the second weighted mean square error of the child nodes associated with each leaf node in the initial random forest model based on the feature weights of each sample feature and the initial prediction results; and determining the node loss of each leaf node in the initial random forest model based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node. The step of determining the node loss of each leaf node in the initial random forest model based on the first weighted mean square error of each leaf node and the second weighted mean square error of each child node includes: in, The first weighted mean square error of the leaf node. It is the left child node of the leaf node. This is the right child node of the leaf node. For the The number of samples, For the The number of samples, For the The second weighted mean square error, For the The method for determining the second weighted mean square error of each leaf node and the second weighted mean square error of the child nodes associated with each leaf node includes: in, Let N be the number of samples at node N. For the predicted value of sample data i, The sample weights are the features of the j-th sample.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images