Method, device and medium for predicting remaining life of battery

Abstract

This invention relates to the field of battery life prediction technology, specifically providing a method, apparatus, and medium for predicting the remaining life of a battery, aiming to solve the problem of inaccurate prediction of the remaining battery life. To this end, the method includes acquiring the battery's full life-cycle operating parameters and dividing these parameters into training and test sets; selecting highly correlated features from the training set to establish a training set health model; and calculating the actual health HI of the test set based on its features. Test And based on the health model of the training set, predict the predicted health HI of the test set. Pred The two models are compared to select the K closest battery health models; the final lifetime TL of these K closest battery health models is obtained, and the predicted lifetime TL of the test set batteries is calculated. Pred Predicted remaining lifetime RUL = TL Pred -TL Curr , among which, TL Curr This method aims to accurately predict the remaining lifespan of a test cell.

Description

Technical Field

[0001] This invention relates to the field of battery life prediction technology, and specifically provides a method, apparatus and medium for predicting the remaining life of a battery. Background Technology

[0002] Because of the complex operating conditions of batteries, even new batteries have certain deviations in their condition after leaving the factory. Therefore, it is very important to calculate the health status of each battery pack and predict its remaining lifespan.

[0003] Existing technologies often predict battery life by measuring voltage and current values ​​during operation. However, due to varying operating conditions, current and voltage are constantly changing, there is no constant current and constant voltage process, and temperature varies with the vehicle's operating environment and time. Furthermore, the state of charge (SOC) changes randomly. As a result, this method has many problems in terms of accuracy.

[0004] Accordingly, there is a need in the art for a new method for predicting the remaining battery life to address the above problems. Summary of the Invention

[0005] To overcome the above-mentioned defects, the present invention is proposed to provide a solution or at least a partial solution to the technical problem of inaccurate prediction of battery remaining life.

[0006] In a first aspect, the present invention provides a method for predicting the remaining lifespan of a battery. The method includes acquiring the battery's full lifespan operating parameters, dividing these parameters into a training set and a test set, selecting highly correlated features from the training set to establish a training set health model, and calculating the actual health of the test set based on the features of the test set. And predict the health of the test set based on the health model of the training set. The two models are compared to select the K closest battery health models, and the final lifespan of the K closest battery health models is obtained. Calculate the predicted lifetime TL of the test cell. Pred Predicted remaining lifetime RUL=TL Pred -TL Curr Among them, TL Curr To test the current lifespan of the collector cell.

[0007] In one technical solution of the aforementioned method for predicting the remaining lifespan of a battery, the steps of "obtaining the battery's full lifespan operating parameters and dividing them into training and testing sets" specifically include: obtaining the battery capacity, DC internal resistance, cumulative ampere-hours, and equivalent cycle count; calculating the capacity decay factor based on the capacity decay and the internal resistance increase factor based on the increase in DC internal resistance; establishing a feature dataset using the battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, capacity decay factor, and internal resistance increase factor as the battery's full lifespan operating parameters and performing data preprocessing on it; and dividing the feature dataset into training and testing sets.

[0008] In one technical solution of the above-mentioned method for predicting the remaining battery life, the step of "calculating the capacity decay factor based on the battery capacity decay" specifically includes:

[0009] The formula for calculating the capacity decay factor is:

[0010]

[0011] in, It is the capacity decay factor. For the current battery capacity, The rated capacity of the battery

[0012] In one technical solution of the above-mentioned method for predicting the remaining battery life, the step of "calculating the internal resistance increase factor based on the increase in DC internal resistance" specifically includes:

[0013] The formula for calculating the internal resistance increase factor is as follows:

[0014]

[0015] in, To increase the internal resistance factor, The current DC internal resistance, This is the rated DC internal resistance.

[0016] In one technical solution of the above-mentioned method for predicting remaining battery life, the step of "preprocessing the feature dataset" specifically includes:

[0017] Through linear transformation The feature dataset is normalized, where x is the feature value. and These are the minimum and maximum values ​​of the feature values ​​in the training set, respectively.

[0018] In one technical solution of the above-mentioned method for predicting remaining battery life, the step of "selecting highly correlated features from the training set to establish a training set health model" specifically includes:

[0019] Calculate the Pearson correlation coefficient between each feature of the training set and the equivalent number of iterations;

[0020] Features with a relevance greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature values ​​Z. Machine learning methods are then used to calculate the health score to be tested using the training set. and build The linear relationship between the number of loops and the equivalent number of iterations.

[0021] In one technical solution of the aforementioned method for predicting remaining battery life, "features with a correlation greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature value Z, and machine learning methods are used to calculate the health score to be tested using the training set." and build The steps for establishing a linear relationship between the number of iterations and the equivalent number of iterations specifically include:

[0022] A multiple linear regression model is constructed by selecting features with a correlation greater than a set threshold C1. ,in, Y represents the training health score, with a value ranging from [0,1]. , The selected feature values ​​are n, where n is the number of features. , is the weight value for the multiple linear regression, and b is the intercept;

[0023] Parameters are obtained using machine learning based on the training set. ;

[0024] According to the formula Calculate the health status of the test subject ;

[0025] Build Multinomial regression model with equivalent number of iterations , where y is x is the equivalent loop count. The order of a polynomial. These are the model parameters.

[0026] In one technical solution of the aforementioned method for predicting remaining battery life, the step of "calculating the Pearson correlation coefficient between features in the training set and the equivalent number of cycles" specifically includes:

[0027] Use the battery's full lifecycle operating parameters in the training set as aging characteristics;

[0028] The Pearson correlation coefficient is calculated based on aging characteristics and health status, using the following formula:

[0029]

[0030] in, The value of is [-1, 1]. , This is a characteristic of battery aging. For health, This represents the average value of battery aging characteristics. To train the mean health score; features with low correlation coefficients are removed.

[0031] In one technical solution of the aforementioned method for predicting battery remaining life, in "constructing..." The steps preceding the "multinomial regression model with equivalent iteration count" also include:

[0032] Health status to be tested The filtering process includes moving average filtering.

[0033] In one technical solution of the aforementioned method for predicting remaining battery life, "the test health level is calculated based on the characteristics of the test set". And predict the health of the test set based on the health model of the training set. The specific steps of "comparing the two to select the K closest battery health models" include:

[0034] According to the formula Calculate the actual health of the test set, where X is a feature in the test set;

[0035] According to the formula Calculate the predicted health of the test set, where x is the equivalent number of loops for the test set;

[0036] According to the formula The difference between the predicted health score and the actual health score of the test set is calculated as the residual.

[0037] According to the formula Calculate the similarity vector and apply the formula. The average similarity score is calculated to measure the degree of similarity. The models are then ranked according to their similarity scores, and the K highest-scoring battery health models are selected. Let be a q-dimensional vector. For the test set There are 16 batteries, where j is the j-th battery model in the training set.

[0038] In one technical solution of the above-mentioned method for predicting the remaining battery life, the method further includes, if the scores calculated using the similarity formula cannot distinguish between the two, then according to the formula: Calculate the similarity vector and apply the formula: The average similarity is calculated as a similarity score to measure the degree of similarity.

[0039] In one technical solution of the above-mentioned method for predicting the remaining battery life, the method further includes determining the actual health of the test set. Then, it is processed by moving average filtering.

[0040] In one technical solution of the aforementioned method for predicting remaining battery life, "the final lifespan of the K closest battery health models is obtained" Calculate the predicted lifetime of the test cell. The specific steps include:

[0041] Test the predicted lifetime of the collector cell The calculation formula is:

[0042] ,

[0043] ,

[0044] ;

[0045] in, The weight corresponding to the battery ranked i in the similarity score. = 1, 2… k, The lifespan at the end of the life of the battery ranked i in similarity score is the equivalent number of cycles or the cumulative ampere-hours.

[0046] In a second aspect, the present invention provides a battery remaining life prediction device, comprising a processor and a memory, the memory being adapted to store a plurality of program codes, the program codes being adapted to be loaded by the processor and run the aforementioned battery remaining life prediction method.

[0047] In a third aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the aforementioned method for predicting the remaining battery life.

[0048] Option 1. A method for predicting the remaining life of a battery, comprising:

[0049] Obtain the battery's full lifecycle operating parameters and divide these parameters into training and testing sets;

[0050] A training set health model is established by selecting highly correlated features from the training set.

[0051] Calculate the actual health of the test set based on its characteristics. And predict the health of the test set based on the health model of the training set. The two models are compared to select the K closest battery health models.

[0052] Obtain the final lifespan of the K closest battery health models. Calculate the predicted lifetime TL of the test cell. Pred ;

[0053] Predicted remaining lifetime RUL=TL Pred -TL Curr , among which, TL Curr To test the current lifespan of the collector cell.

[0054] Option 2. According to the battery remaining life prediction method described in Option 1, the steps of "obtaining the battery's full life cycle operating parameters and dividing the battery's full life cycle operating parameters into training and testing sets" specifically include:

[0055] Obtain battery capacity, DC internal resistance, cumulative ampere-hours, and equivalent cycle count;

[0056] Calculate the capacity decay factor based on the decrease in battery capacity, and calculate the internal resistance increase factor based on the increase in DC internal resistance.

[0057] A feature dataset was established using battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, capacity decay factor, and internal resistance increase factor as operating parameters for the entire battery life cycle, and the data was preprocessed.

[0058] The feature dataset is divided into a training set and a test set.

[0059] Option 3. According to the battery remaining life prediction method described in Option 2, the step of "calculating the capacity degradation factor based on the battery capacity degradation" specifically includes:

[0060] The formula for calculating the capacity decay factor is:

[0061]

[0062] in It is the capacity decay factor. For the current battery capacity, This refers to the battery's rated capacity.

[0063] Option 4. According to the battery remaining life prediction method described in Option 2, the specific steps of "calculating the internal resistance increase factor based on the increase in DC internal resistance" include:

[0064] The formula for calculating the internal resistance increase factor is as follows:

[0065]

[0066] in, To increase the internal resistance factor, The current DC internal resistance, This is the rated DC internal resistance.

[0067] Option 5. According to the battery remaining life prediction method described in Option 2, the step of "preprocessing the feature dataset" specifically includes:

[0068] Through linear transformation The feature dataset is normalized, where x is the feature value. and These are the minimum and maximum values ​​of the feature values ​​in the training set, respectively.

[0069] Option 6. According to the battery remaining life prediction method described in Option 1, the specific steps of "selecting highly correlated features in the training set to establish a training set health model" include:

[0070] Calculate the Pearson correlation coefficient between each feature of the training set and the equivalent number of iterations;

[0071] Features with a relevance greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature values ​​Z. Machine learning methods are then used to calculate the health score to be tested using the training set. and build The linear relationship between the number of loops and the equivalent number of iterations.

[0072] Option 7. Based on the battery remaining life prediction method described in Option 6, “features with a correlation greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature value Z, and machine learning methods are used to calculate the health score to be tested using the training set.” and build The steps for establishing a linear relationship between the number of iterations and the equivalent number of iterations specifically include:

[0073] Features with a correlation greater than a set threshold C1 are selected to construct a multiple linear regression model. ,in, Y represents the training health score, with a value ranging from [0,1]. , The selected feature values ​​are n, where n is the number of features. , is the weight value for the multiple linear regression, and b is the intercept;

[0074] Parameters are obtained using machine learning based on the training set. ;

[0075] According to the formula Calculate the health status of the test subject ;

[0076] Build Multinomial regression model with equivalent number of iterations , where y is x is the equivalent loop count. Let be the order of the polynomial. These are the model parameters.

[0077] Option 8. According to the battery remaining life prediction method described in Option 7, the specific steps of "calculating the Pearson correlation coefficient between the features in the training set and the equivalent number of cycles" include:

[0078] Use the battery's full lifecycle operating parameters in the training set as aging characteristics;

[0079] The Pearson correlation coefficient is calculated based on aging characteristics and health status, using the following formula:

[0080]

[0081] in, The value of is [-1, 1]. , This is a characteristic of battery aging. For health, This represents the average value of battery aging characteristics. To train the average health score;

[0082] Features with low relevance are removed based on the correlation coefficient.

[0083] Option 9. Based on the battery remaining life prediction method described in Option 8, in "Constructing..." The steps preceding the "multinomial regression model with equivalent iteration count" also include:

[0084] Health status to be tested The filtering process includes moving average filtering.

[0085] Option 10. Based on the battery remaining life prediction method described in Option 9, "calculate the test health status based on the characteristics of the test set". And predict the health of the test set based on the health model of the training set. The specific steps of "comparing the two to select the K closest battery health models" include:

[0086] According to the formula Calculate the actual health of the test set, where X is a feature in the test set;

[0087] According to the formula Calculate the predicted health of the test set, where x is the equivalent number of loops for the test set;

[0088] According to the formula The difference between the predicted health score and the actual health score of the test set is calculated as the residual.

[0089] According to the formula Calculate the similarity vector and apply the formula. The average similarity score is calculated to measure the degree of similarity. The models are then ranked according to their similarity scores, and the K highest-scoring battery health models are selected. Let be a q-dimensional vector. For the test set There are 16 batteries, where j is the j-th battery model in the training set.

[0090] Option 11. The method for predicting the remaining battery life according to Option 10 further includes:

[0091] If the scores obtained using the similarity formula cannot distinguish between the two, then the formula is applied as follows: Calculate the similarity vector and apply the formula: The average similarity is calculated as a similarity score to measure the degree of similarity.

[0092] Option 12. The method for predicting the remaining battery life according to Option 10 further includes:

[0093] To determine the actual health of the test set Then, it is processed by moving average filtering.

[0094] Option 13. Based on the battery remaining life prediction method described in Option 12, "obtain the final life of the K closest battery health models". Calculate the predicted lifetime TL of the test cell. Pred The specific steps include:

[0095] Test cell predicted lifetime TL Pred The calculation formula is:

[0096] ,

[0097] ,

[0098] ;

[0099] in, Ranked by similarity score The weight corresponding to the battery, = 1, 2… k, Ranked by similarity score The lifespan at the end of the battery's lifespan is the equivalent number of cycles or the cumulative ampere-hours.

[0100] Option 14. A battery remaining life prediction device, comprising a processor and a memory, the memory being adapted to store a plurality of program codes, the program codes being adapted to be loaded by the processor and executed to perform the battery remaining life prediction method as described in any one of Options 1 to 13.

[0101] Option 15. A computer-readable storage medium storing a plurality of program codes adapted to be loaded and run by a processor to perform a method for predicting the remaining battery life as described in any one of Options 1 to 13.

[0102] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0103] In implementing the technical solution of this invention, the data is first preprocessed by obtaining the battery's full life cycle operating parameters through databases, and then normalized. The operating parameters are divided into training set data and test set data. Next, the training set data is modeled using multinomial fitting, and then the test set data is predicted using the training set model. Finally, the remaining lifespan is calculated based on the similarity score. This method can accurately predict the remaining lifespan of the battery through machine learning.

[0104] The system obtains battery capacity, DC internal resistance, cumulative ampere-hours, and equivalent cycle count. It calculates the capacity decay factor based on the capacity decay and the internal resistance increase factor based on the increase in DC internal resistance. The dataset established by using these parameters as the battery's operating characteristic values ​​accurately and comprehensively reflects the battery's operating status and provides comprehensive data support for subsequent calculations of battery life.

[0105] Selecting features with a correlation coefficient greater than a set threshold C1 to construct a multiple linear regression model can eliminate features with low correlation, reduce computation, and improve modeling efficiency.

[0106] Filtering the battery health data, such as using a moving average, can eliminate some of the noise.

[0107] The K battery health models with the highest similarity scores are selected, and different weight values ​​are assigned to them according to their scores, so as to obtain a highly accurate battery life prediction and thus a highly accurate remaining life. Attached Figure Description

[0108] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0109] Figure 1 This is a schematic flowchart of the main steps of a method for predicting the remaining battery life according to an embodiment of the present invention;

[0110] Figure 2 This is a flowchart illustrating the specific steps of data processing in a method for predicting battery remaining life according to an embodiment of the present invention.

[0111] Figure 3 This is a schematic diagram of the fitting curve of equivalent cycle number and health status according to an embodiment of the present invention. Detailed Implementation

[0112] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0113] In the description of this invention, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0114] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a method for predicting the remaining battery life according to an embodiment of the present invention. Figure 1 As shown, the method for predicting the remaining battery life in this embodiment of the invention mainly includes the following steps S11-S15.

[0115] Step S11: Obtain the battery's full life cycle operating parameters and divide the battery's full life cycle operating parameters into training set and test set.

[0116] In this embodiment, the battery's full lifecycle operating parameters can be obtained from a database, from the vehicle's battery management system, or through other effective means. The obtained data is divided into two parts: one part is used as a training set to build a model, and the other part is used as a test set to predict the remaining lifespan and verify the model's correctness.

[0117] See appendix Figure 2 , Figure 2 This is a flowchart illustrating the specific steps of data processing in a method for predicting battery remaining life according to an embodiment of the present invention. Figure 2 As shown, the steps of “obtaining the battery’s full life cycle operating parameters and dividing the battery’s full life cycle operating parameters into training set and test set” specifically include the following steps S111-S114.

[0118] Step S111: Obtain battery capacity, DC internal resistance, cumulative ampere-hours, and equivalent cycle count.

[0119] In this embodiment, the battery has many operating parameters, including, for example, equivalent temperature, equivalent current, equivalent voltage, etc.

[0120] Step S112: Calculate the capacity decay factor based on the decrease in battery capacity, and calculate the internal resistance increase factor based on the increase in DC internal resistance.

[0121] In one embodiment, the step of "calculating the capacity decay factor based on the battery capacity degradation" specifically includes the following formula for calculating the capacity decay factor:

[0122]

[0123] in, For the current battery capacity, This refers to the battery's rated capacity.

[0124] In one implementation, the step of "calculating the internal resistance increase factor based on the increase in DC internal resistance" specifically includes:

[0125] The formula for calculating the internal resistance increase factor is as follows:

[0126]

[0127] in, To increase the internal resistance factor, The current DC internal resistance, This is the rated DC internal resistance.

[0128] Specifically, new features are constructed based on the battery capacity decay and the increase in DC internal resistance, namely the capacity decay factor and the internal resistance increase factor. These two calculated features, together with the original operating parameters, form a feature dataset.

[0129] Step S113: Establish a feature dataset by using battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, capacity decay factor, and internal resistance increase factor as operating parameters for the entire battery life cycle, and perform data preprocessing on the dataset.

[0130] In one implementation, the step of "preprocessing the feature dataset" specifically includes: through linear transformation The feature dataset is normalized, where x is the feature value. and These are the minimum and maximum values ​​of the feature values ​​in the training set, respectively.

[0131] Specifically, after the feature dataset is established, it is normalized to the range of 0-1. First, the training set is normalized by using the maximum and minimum values ​​of each feature value through a linear transformation. Then, the features of the test set are normalized using the maximum and minimum values ​​of the training set features.

[0132] Step S114: Divide the feature dataset into a training set and a test set.

[0133] Specifically, a preferred approach is to use the k-fold cross-validation method to divide the feature dataset into training and test sets. The k-fold method is a commonly used method known in this field. There are various ways to divide the training and test sets, and the training and test sets can even be converted to each other, as long as the testing requirements are met.

[0134] By combining steps S111-S114 above, the data processing can be completed, and training and test sets can be established for later use.

[0135] Step S12: Select highly correlated features from the training set to establish a training set health model.

[0136] In this embodiment, in order to reduce the amount of computation and improve the accuracy of model construction, features with high correlation in the training set are preferably selected.

[0137] In one implementation, the step of "selecting highly correlated features from the training set to establish a training set health model" specifically includes: calculating the Pearson correlation coefficient between each feature in the training set and the equivalent number of iterations; selecting features with a correlation greater than a set threshold C1; constructing a linear relationship between the training health Y and the selected feature values ​​Z; and using machine learning methods to calculate the health score to be tested using the training set. and build The linear relationship between the number of loops and the equivalent number of iterations.

[0138] In one implementation, "features with a relevance greater than a set threshold C1 are selected, a linear relationship between the training health score Y and the selected feature value Z is constructed, and machine learning methods are used to calculate the health score to be tested using the training set." and build The steps for establishing a linear relationship between the number of iterations and the equivalent number of iterations specifically include selecting features with a correlation greater than a set threshold C1 to construct a multiple linear regression model. ,in, Y represents the training health score, with a value ranging from [0,1]. , The selected feature values ​​are n, where n is the number of features. , Here, b represents the weights for the multiple linear regression, and b is the intercept; parameters are obtained using machine learning based on the training set. According to the formula Calculate the health status of the test subject ; build Multinomial regression model with equivalent number of iterations , where y is x is the equivalent loop count. Let be the order of the polynomial. These are the model parameters.

[0139] In one implementation, the step of "calculating the Pearson correlation coefficient between features in the training set and the equivalent number of cycles" specifically includes: using the battery's full life-cycle operating parameters in the training set as aging features; and calculating the Pearson correlation coefficient based on the aging features and health status, using the following formula:

[0140]

[0141] in, The value of is [-1, 1]. , This is a characteristic of battery aging. For health, This represents the average value of battery aging characteristics. This is the average health score for training.

[0142] Specifically, the training set is modeled, using battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, equivalent temperature, equivalent current, capacity decay factor, and internal resistance increase factor as aging features. These features are combined with health status, and the correlation coefficient is calculated using the Pearson correlation coefficient formula. This correlation coefficient is then compared with a pre-set threshold C1. The threshold C1 can preferably be set to values ​​such as 0.2 or 0.3. After selecting the relevant features, a multiple linear regression model is used... We can use machine learning methods to obtain what we need. This refers to the weights in the multiple linear regression model. Therefore, the multiple linear regression model can be... It is applied to the test set to calculate the actual health of the test set under the training set multiple linear regression model.

[0143] In one implementation, in "building" Before the step of "multinomial regression model with equivalent number of iterations", the following steps are also included: the health status to be measured. The filtering process includes moving average filtering.

[0144] Specifically, filtering can reduce parameter noise and improve the parameter's anti-interference ability.

[0145] Step S13: Calculate the actual health of the test set based on its characteristics. And predict the health of the test set based on the health model of the training set. We compare the two models to select the K closest battery health models.

[0146] In this embodiment, the actual health of the test set and predicting health Based on different formulas, the metalinear regression model has already been used for the training set. The calculated health score of the test set is considered the actual health score. .

[0147] See appendix Figure 3 , Figure 3 This is a schematic diagram of the fitting curve of equivalent cycle number and health status according to an embodiment of the present invention.

[0148] like Figure 3 As shown, in one implementation, "the test health score is calculated based on the characteristics of the test set". And predict the health of the test set based on the health model of the training set. The specific steps of "comparing the two to select the K closest battery health models" include:

[0149] According to the formula Calculate the actual health of the test set, where X is a feature in the test set;

[0150] According to the formula Calculate the predicted health of the test set, where x is the equivalent number of loops for the test set;

[0151] According to the formula The difference between the predicted health score and the actual health score of the test set is calculated as the residual.

[0152] According to the formula Calculate the similarity vector and apply the formula. The average similarity score is calculated to measure the degree of similarity. The models are then ranked according to their similarity scores, and the K highest-scoring battery health models are selected. Let be a q-dimensional vector. For the test set There are 16 batteries, where j is the j-th battery model in the training set.

[0153] Specifically, in the previous step, the calculation was already performed based on the training set. Based on the numerical value, the features of the test set are substituted into the formula to obtain the actual health score of the test set. Then, using the selected equivalent number of loops as the x-axis (X-value), the predicted health score of the test set is obtained according to the above formula. For example... Figure 3 As shown, HI represents health level, and cycle represents the equivalent cycle count. The correspondence between the equivalent cycle count and health level is intuitively apparent, and the vertical dashed line in the figure represents the equivalent cycle count and corresponding health level in the current state. The residual between the predicted and actual health levels is obtained by taking the absolute value of the difference between the actual and predicted health levels; clearly, the smaller the residual, the smaller the error. Subsequently, the similarity is calculated and averaged. The models are then sorted according to their scores, and the K highest-scoring battery health models are selected. K is a manually assigned value that can be set as needed.

[0154] In one implementation, if the scores calculated using the similarity formula are indistinguishable, then according to the formula: Calculate the similarity vector and apply the formula: The average similarity is calculated as a similarity score to measure the degree of similarity.

[0155] Specifically, when the results calculated using the similarity calculation formula in the previous step are too similar, the formula in this embodiment can be used for calculation, which provides a more refined similarity score and higher contrast.

[0156] In one implementation, the actual health of the test set is determined. Then, it is processed by moving average filtering.

[0157] Specifically, the actual health status here can still be filtered to reduce parameter noise and improve the parameter's anti-interference ability.

[0158] Step S14: Obtain the final lifespan of the K closest battery health models. Calculate the predicted lifetime TL of the test cell. Pred .

[0159] In this embodiment, different weights are set according to the similarity value. Calculate the predicted lifetime TL of the test cell. Pred .

[0160] In one implementation, "obtain the final lifespan of the K closest battery health models" Calculate the predicted lifetime TL of the test cell. Pred The specific steps include:

[0161] Test cell predicted lifetime TL Pred The calculation formula is:

[0162] ,

[0163] ,

[0164] ;

[0165] in, Ranked by similarity score The weight corresponding to the battery, = 1, 2… k, Ranked by similarity score The lifespan at the end of the battery's lifespan is the equivalent number of cycles or the cumulative ampere-hours.

[0166] Specifically, the final lifespan of the K highest-rated battery models from the training set is obtained. Then, weights are assigned according to the scores, with higher scores having greater weights. Each weight is equal to the average similarity score.

[0167] Step S15: Predict remaining lifetime RUL=TL Pred -TL Curr Among them, TL Curr To test the current lifespan of the collector cell.

[0168] In this embodiment, the difference between the calculated predicted remaining lifespan and the current lifespan that has already been experienced is the predicted remaining lifespan of the battery.

[0169] Based on steps S11-S15 above, the data is first preprocessed by obtaining the battery's full life cycle operating parameters through databases, battery management systems, etc., and then normalizing them. The operating parameters are divided into training set data and test set data. Next, multinomial fitting is used to model the training set data, and then the training set model is used to predict the test set data. Finally, the remaining lifespan is calculated based on the similarity score. This method can accurately predict the remaining lifespan of the battery through machine learning.

[0170] By acquiring battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, equivalent temperature, and equivalent current, the capacity decay factor is calculated based on the capacity decay, and the internal resistance increase factor is calculated based on the increase in DC internal resistance. The dataset established by using the above parameters as the battery's operating characteristic values ​​accurately and comprehensively reflects the battery's operating status, providing comprehensive data support for subsequent prediction calculations of battery life.

[0171] Selecting features with correlation coefficients greater than a set threshold C1 to construct a multiple linear regression model can eliminate features with low correlation, reduce computation, and improve modeling efficiency. The threshold C1 is preferably 0.2.

[0172] Filtering the battery health data, such as using a moving average, can eliminate some of the noise.

[0173] The K battery health models with the highest similarity scores are selected, and different weight values ​​are assigned to them according to their scores, so as to obtain a highly accurate battery life prediction and thus a highly accurate remaining life.

[0174] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0175] Those skilled in the art will understand that all or part of the processes in the methods of any of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0176] Furthermore, the present invention also provides a battery remaining life prediction device. In a control device embodiment according to the present invention, the battery remaining life prediction device includes a processor and a storage device. The storage device can be configured to store a program for executing the battery remaining life prediction method of the above-described method embodiments. The processor can be configured to execute the program in the storage device, which includes, but is not limited to, a program for executing the battery remaining life prediction method of the above-described method embodiments. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The control device can be a control device device comprising various electronic devices.

[0177] Furthermore, the present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program for performing the battery remaining life prediction method of the above-described method embodiments. This program can be loaded and run by a processor to implement the battery remaining life prediction method. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0178] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device of the present invention, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, hardware, or a combination of software and hardware. Therefore, the number of modules shown in the figures is merely illustrative.

[0179] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of the present invention; therefore, the technical solutions after splitting or combining will fall within the protection scope of the present invention.

[0180] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for predicting the remaining lifespan of a battery, characterized in that, include: Obtain the battery's full lifecycle operating parameters and divide these parameters into training and testing sets; A training set health model is established by selecting highly correlated features from the training set. Calculate the actual health of the test set based on its characteristics. And predict the health of the test set based on the health model of the training set. The two models are compared to select the K closest battery health models. Obtain the final lifespan of the K closest battery health models. Calculate the predicted lifetime TL of the test cell. Pred ; Predicted remaining lifetime RUL=TL Pred -TL Curr , among which, TL Curr To test the current lifespan of the collector cell; Among them, "calculate the actual health of the test set based on its characteristics" And predict the health of the test set based on the health model of the training set. The comparison of the two models to select the K closest battery health models includes: According to the formula Calculate the actual health of the test set, where, n is the number of features. Here, X represents the weights for the multiple linear regression, and X represents the features in the test set. According to the formula Calculate the predicted health score for the test set, where, Let be the order of the polynomial. Here, x represents the model parameters, and x represents the equivalent number of iterations on the test set. According to the formula The difference between the predicted health score and the actual health score of the test set is calculated as the residual. According to the formula Calculate the similarity vector and apply the formula. The average similarity score is calculated to measure the degree of similarity. The models are then ranked according to their similarity scores, and the K highest-scoring battery health models are selected. Let be a q-dimensional vector. For the test set There are 16 batteries, where j is the j-th battery model in the training set.

2. The method for predicting the remaining battery life according to claim 1, characterized in that, The steps of "obtaining the battery's full lifecycle operating parameters and dividing these parameters into training and testing sets" specifically include: Obtain battery capacity, DC internal resistance, cumulative ampere-hours, and equivalent cycle count; Calculate the capacity decay factor based on the decrease in battery capacity, and calculate the internal resistance increase factor based on the increase in DC internal resistance. A feature dataset was established using battery capacity, DC internal resistance, cumulative ampere-hours, equivalent cycle count, capacity decay factor, and internal resistance increase factor as operating parameters for the entire battery life cycle, and the data was preprocessed. The feature dataset is divided into a training set and a test set.

3. The method for predicting the remaining battery life according to claim 2, characterized in that, The steps for "calculating the capacity degradation factor based on the battery capacity degradation" specifically include: The formula for calculating the capacity decay factor is: in It is the capacity decay factor. For the current battery capacity, This refers to the battery's rated capacity.

4. The method for predicting the remaining battery life according to claim 2, characterized in that, The specific steps for "calculating the internal resistance increase factor based on the increase in DC internal resistance" include: The formula for calculating the internal resistance increase factor is as follows: in, To increase the internal resistance factor, The current DC internal resistance, This is the rated DC internal resistance.

5. The method for predicting the remaining battery life according to claim 2, characterized in that, The specific steps involved in "preprocessing the feature dataset" include: Through linear transformation The feature dataset is normalized, where x is the feature value. and These are the minimum and maximum values ​​of the feature values ​​in the training set, respectively.

6. The method for predicting the remaining battery life according to claim 1, characterized in that, The specific steps involved in "selecting highly correlated features from the training set to build a training set health model" include: Calculate the Pearson correlation coefficient between each feature of the training set and the equivalent number of iterations; Features with a relevance greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature values ​​Z. Machine learning methods are then used to calculate the health score to be tested using the training set. and build The linear relationship between the number of loops and the equivalent number of iterations.

7. The method for predicting the remaining battery life according to claim 6, characterized in that, "Features with a relevance greater than a set threshold C1 are selected to construct a linear relationship between the training health score Y and the selected feature values ​​Z. Machine learning methods are then used to calculate the health score to be tested using the training set." and build The steps for establishing a linear relationship between the number of iterations and the equivalent number of iterations specifically include: Features with a correlation greater than a set threshold C1 are selected to construct a multiple linear regression model. ,in, , For training health, the value range is [0,1]. , The selected feature values ​​are n, where n is the number of features. , b represents the weights of the multiple linear regression, and b is the intercept. Parameters are obtained using machine learning based on the training set. ; According to the formula Calculate the health status of the test subject , where X is a feature in the test set; Build Multinomial regression model with equivalent number of iterations , where y is x is the equivalent loop count. Let be the order of the polynomial. These are the model parameters.

8. The method for predicting the remaining battery life according to claim 7, characterized in that, The steps for "calculating the Pearson correlation coefficient between features in the training set and the equivalent number of iterations" specifically include: Use the battery's full lifecycle operating parameters in the training set as aging characteristics; The Pearson correlation coefficient is calculated based on aging characteristics and health status, using the following formula: in, The value of is [-1, 1]. , This is a characteristic of battery aging. For health, This represents the average value of battery aging characteristics. To train the average health score; Features with low relevance are removed based on the correlation coefficient.

9. The method for predicting the remaining battery life according to claim 8, characterized in that, In "building" The steps preceding the "multinomial regression model with equivalent iteration count" also include: Health status to be tested The filtering process includes moving average filtering.

10. The method for predicting the remaining battery life according to claim 1, characterized in that, Also includes: Or according to the formula: Calculate the similarity vector and apply the formula: The average similarity is calculated as a similarity score to measure the degree of similarity.

11. The method for predicting the remaining battery life according to claim 1, characterized in that, Also includes: To determine the actual health of the test set Then, it is processed by moving average filtering.

12. The method for predicting the remaining battery life according to claim 1, characterized in that, "Obtain the final lifespan of the K closest battery health models" Calculate the predicted lifetime TL of the test cell. Pred The specific steps include: Test cell predicted lifetime TL Pred The calculation formula is: ， ， ； in, Ranked by similarity score The weight corresponding to the battery, = 1, 2… k, Ranked by similarity score The lifespan at the end of the battery's lifespan is the equivalent number of cycles or the cumulative ampere-hours.

13. A device for predicting the remaining life of a battery, comprising a processor and a memory, the memory being adapted to store multiple lines of program code, characterized in that, The program code is adapted to be loaded and run by the processor to perform the method for predicting the remaining battery life as claimed in any one of claims 1 to 12.

14. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the method for predicting the remaining battery life as described in any one of claims 1 to 12.

Images