Lithium ion battery direct current internal resistance-based battery health state prediction method and system

Abstract

The invention discloses a lithium ion battery DC internal resistance-based battery health state prediction method and system, and relates to the technical field of battery health state prediction. The method comprises the following steps: collecting pulse voltage data of a whole life cycle of a single lithium ion battery, and preprocessing the pulse voltage data; extracting a battery voltage rebound characteristic curve according to the preprocessed pulse voltage data; according to the extracted battery voltage rebound characteristic curve, the current battery health state is evaluated, and an evaluation result is output; and predicting the battery life according to the current battery health state evaluation result, and outputting a prediction result. According to the method and the system, the process of predicting the health state of the battery can be effectively simplified in the process of predicting the health state of the battery, the method and the system can be widely applied to actual situations, and the precision and the efficiency of predicting the health state of the battery are effectively improved; and internal aging and safety conditions of the battery are effectively fed back, so that an abnormal battery is timely replaced and handled.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health state prediction, and in particular to a battery health state prediction method and system based on the direct current internal resistance of a lithium-ion battery. Background Art

[0002] Lithium-ion batteries have been widely used in many fields in recent years due to their high energy density, long service life and relatively low cost, and have become an important energy storage device for electric vehicles. However, as the number of charging cycles increases, the battery undergoes irreversible aging, such as reduced battery life and even safety hazards. Therefore, accurate prediction of the battery's state of health (SOH) and remaining useful life (RUL) will bring new technological advancements to the fields of battery manufacturing and BMS optimization. In addition, accurate prediction of the remaining life of the battery will help to achieve battery recycling, improve resource utilization, and reduce the harm of waste batteries to the environment, thereby promoting the green development of the new energy vehicle industry.

[0003] In existing technical research, there are three main types of health diagnosis methods for lithium batteries: physical models, semi-empirical models, and data-driven methods. The goal of physical and semi-empirical models is to reveal the various mechanisms related to battery capacity decay, such as the formation of solid electrolyte interface (SEI) film, lithium deposition, active material consumption, and internal resistance increase. Although these models can successfully predict the life of the battery to a certain extent, in actual applications, the charging and discharging behavior of the battery is often very complex and irregular. Various cycle conditions (such as charge and discharge rates, cycle temperatures, etc.) cause different aging behaviors of the battery. Therefore, it is very challenging to construct a physical model or semi-empirical model suitable for the actual behavior of users to accurately predict battery health diagnosis and safety warnings.

[0004] With the rise of artificial intelligence technology, data-driven methods based on machine learning and neural networks are widely used in battery health diagnosis. This data-driven method does not require a deep understanding of the electrochemical principles of the battery. It only needs a large amount of data to train the model to achieve the prediction function. The data-driven method has high requirements for the aging characteristic factors of the battery. The more the characteristic factors can reflect the aging law of the battery, the higher the accuracy. In current research, commonly used characteristic factors include charge and discharge curves, incremental capacity curves (IC), electrochemical impedance spectroscopy (EIS), etc.

[0005] However, there are great limitations when using the above methods to predict the health status of batteries. The charge and discharge curve must fully record the voltage, current and time data in each charge and discharge cycle. The process is cumbersome and requires high calculations. The incremental capacity curve analysis requires high-frequency data collection and differential calculations, which is not conducive to widespread application in actual situations. The EIS measurement process is complicated and the equipment is expensive, which increases the test cost and is not conducive to large-scale promotion. The above methods result in low accuracy and efficiency in battery health status prediction. Summary of the invention

[0006] In order to overcome the problem that the existing battery health status prediction technology has the problem that the battery health status prediction process is relatively cumbersome, which is not conducive to wide application in actual situations, thereby resulting in low accuracy and efficiency of battery health status prediction, the present invention proposes a battery health status prediction method and system based on the DC internal resistance of a lithium-ion battery, which can effectively simplify the battery health status prediction process, meet the needs of wide application in actual situations, and thus effectively improve the accuracy and efficiency of battery health status prediction.

[0007] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0008] A method for predicting the health status of a battery based on the DC internal resistance of a lithium-ion battery, the method comprising the following steps:

[0009] Collecting pulse voltage data of a single lithium-ion battery throughout its life cycle and preprocessing the pulse voltage data;

[0010] Extracting the battery voltage rebound characteristic curve according to the pre-processed pulse voltage data;

[0011] According to the extracted battery voltage rebound characteristic curve, the current battery health status is evaluated and the evaluation result is output;

[0012] Predict the battery life based on the current battery health status assessment result and output the prediction result.

[0013] In the above technical scheme, preprocessing the collected pulse voltage data can effectively improve the stability and reliability of the data; extracting the battery voltage rebound characteristic curve based on the preprocessed pulse voltage data, that is, applying a pulse current and collecting voltage rebound data, can greatly simplify the experimental process, conduct real-time testing, and reduce instrument costs and measurement costs; evaluating the battery health status based on the extracted battery voltage rebound characteristic curve can effectively simplify the battery health status evaluation process, meet the needs of wide application in actual situations, and effectively predict the battery life based on the current battery health status evaluation results, thereby improving the accuracy and efficiency of battery health status prediction, and effectively feeding back the internal aging and safety conditions of the battery so as to replace and handle abnormal batteries in a timely manner.

[0014] Furthermore, the process of preprocessing the pulse voltage data includes:

[0015] The collected pulse voltage data is uploaded to the lithium battery health database, and the historical data in the battery health database is cleaned.

[0016] Furthermore, the process of cleaning the historical data in the battery health database includes:

[0017] The historical data in the battery health database is sequentially sorted into tables, missing values ​​are filled, rows and columns are split, duplicate data are deleted, data is normalized, and data is integrated to obtain the voltage data and capacity data during the battery pulse discharge;

[0018] The expression for normalizing the historical data in the battery health database is:

[0019]

[0020] Among them, x1 represents the unnormalized voltage data; x max and x min Indicates the maximum and minimum voltage values ​​in different cycles at the same time.

[0021] In the above technical solution, the collected pulse voltage data are sorted into tables, missing values ​​are filled, rows and columns are split, duplicate data are deleted, data is normalized and data is integrated, which can effectively improve the stability and reliability of the data.

[0022] Furthermore, the process of extracting the battery voltage rebound characteristic curve includes:

[0023] For batteries that cycle throughout their life cycle, a complete pulse charging signal is applied in each cycle to extract the battery voltage rebound characteristic curve during the entire pulse charging process;

[0024] Perform parameter identification on the voltage rebound characteristic curve during a pulse charging process to obtain DC internal resistance data, and divide the DC internal resistance data into a training set and a test set;

[0025] The battery capacity data is calculated using the battery cabinet's own algorithm using the ampere-hour integration method.

[0026] In the above technical scheme, a pulse current is applied to collect voltage rebound data, which can greatly simplify the experimental process, conduct tests in real time, and reduce instrument costs and measurement costs; and the DC internal resistance of the battery is reflected by the voltage rebound data, and the experiment can be carried out at any state of charge (SOC) without the need for a complete charging and discharging process, which can effectively simplify the process of battery health status prediction. At the same time, there are differences in the prediction accuracy of different SOCs, and the prediction results of more stable electrochemical states are also relatively stable. In addition, battery health diagnosis based on DC internal resistance can greatly simplify the experimental process, conduct tests in real time, and reduce instrument costs and measurement costs.

[0027] Furthermore, the process of evaluating the current battery health status according to the extracted battery voltage rebound characteristic curve includes:

[0028] Constructing a battery health status assessment model, wherein the battery health status assessment model includes a convolution module, a bidirectional long short-term memory module and an assessment module;

[0029] Inputting the training set into the battery health state assessment model, and setting several rounds of parameter iterative optimization training to optimize the parameters of the battery health state assessment model;

[0030] Use the convolution module to extract the battery health status features in the training set;

[0031] The bidirectional long short-term memory module is used to extract the long-term correlation features in the battery pulse charging and discharging process in the training set;

[0032] The prediction module evaluates the battery health status based on the extracted battery health status features and long-term correlation features, and outputs the evaluation results.

[0033] The prediction results are tested using the test set, and the loss function is set to optimize the parameters of the battery health status assessment model after each round of training. When the preset iterative training round ends or the loss function converges, the trained battery health status assessment model is obtained;

[0034] Among them, the battery health status represents the maximum capacity of the battery's current cycle divided by the battery's initial capacity, and the evaluation result includes the battery's early capacity data.

[0035] Furthermore, the process of extracting long-term correlation features in the battery pulse charging and discharging process in the training set using the bidirectional long short-term memory module includes:

[0036] The bidirectional long short-term memory module includes an input layer, a hidden layer, and an output layer;

[0037] The input layer, hidden layer and output layer are used to extract the long time series in the battery cycle charge and discharge process, and the long-term correlation features are extracted based on the long time series. The expression is:

[0038] f(t)=σ(W g x(t)+W f h(t-1)+b f );

[0039] i(t)=σ(W i x(t)+W i h(t-1)+b i );

[0040] o(t)=σ(W o x(t)+W o h(t-1)+b o );

[0041] C(t)=f(t)⊙C(t-1)+i(t)⊙tanh(W c x(t)+W c h(t-1)+b c );

[0042] y(t)=o(t)⊙tanh(C(t));

[0043] The loss function is set to optimize the parameters of the battery health status assessment model after each round of training:

[0044]

[0045] Where x(t) and y(t) represent the input vector and output vector of the current time step t, respectively; f(t) represents the activation value of the forget gate; i(t) represents the activation value of the input gate; o(t) represents the activation value of the output gate; and Y i , and Represent the true value, predicted value and the average of the true value, R 2 represents the coefficient of determination, and RMSE represents the root mean square error.

[0046] In the above technical scheme, the constructed battery health status assessment model can effectively grasp the changes in pulse voltage data of lithium-ion batteries throughout their life cycle after training with training sets, testing with test sets, and optimizing the loss function. Then, based on the cooperation between the convolution module, the bidirectional long short-term memory module, and the evaluation module, the health status of the current battery is assessed, effectively simplifying the process of battery health status prediction throughout its entire cycle, and can be widely used in actual situations, thereby effectively improving the accuracy and efficiency of battery health status prediction.

[0047] Furthermore, the process of predicting the battery life according to the current battery health status evaluation result includes:

[0048] Predict the battery capacity decay trajectory based on the battery's early capacity data;

[0049] Based on the battery capacity decay trajectory, the battery end-of-life and current cycle number are predicted, and the battery remaining life is predicted based on the battery end-of-life and current cycle number.

[0050] Furthermore, the process of predicting the battery capacity decay trajectory includes:

[0051] Constructing a battery capacity decay trajectory prediction model, wherein the battery capacity decay trajectory prediction model includes a self-attention layer, a feedforward neural network layer, and a battery capacity decay trajectory prediction layer;

[0052] Inputting the training set into the battery capacity decay trajectory prediction model, and setting several rounds of parameter iterative optimization training to optimize the parameters of the battery capacity decay trajectory prediction model;

[0053] The self-attention layer is used to extract the battery capacity features in the training set. The feedforward neural network layer processes the extracted battery capacity features point by point. The battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity features and outputs the prediction results.

[0054] The prediction results are tested using the test set, and the loss function is set to optimize the parameters of the battery capacity decay trajectory prediction model after each round of training. When the preset iterative training round ends or the loss function converges, the trained battery capacity decay trajectory prediction model is obtained.

[0055] Furthermore, the self-attention layer uses the self-attention mechanism to extract the battery capacity features in the training set, expressed as:

[0056]

[0057] The feedforward neural network layer processes the output of the self-attention layer point by point, and the expression is:

[0058] FFN(x)=max(0,xW1+b1)W2+b2;

[0059] The battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity characteristics. The expression is:

[0060] Output = softmax(W out DecoderOutput+b iut );

[0061] The remaining battery life is predicted based on the battery end-of-life and the current number of cycles. The expression is:

[0062] RUL = battery life - current number of charge and discharge cycles of the battery;

[0063] Where Q is the query matrix, K is the key matrix, and V is the value matrix; d k is the dimension of the key vector, x represents the input feature, W1 and W2 represent the weight matrices of the feedforward neural network, b1 and b2 represent the bias vectors of the feedforward neural network, RUL is the remaining battery life finally calculated, and RUL is the remaining battery life finally calculated.

[0064] In the above technical scheme, the constructed battery capacity decay trajectory prediction model can effectively grasp the changes in pulse voltage data of lithium-ion batteries throughout their life cycle after training with training sets, testing with test sets, and optimizing the loss function. Then, based on the mutual cooperation between the self-attention layer, the feedforward neural network layer, and the battery capacity decay trajectory prediction layer, the capacity and life of the battery can be predicted, which effectively simplifies the process of battery health status prediction and can be widely used in actual situations, thereby effectively improving the accuracy and efficiency of battery health status prediction.

[0065] A battery health status prediction system based on direct current internal resistance of a lithium-ion battery, the system comprising:

[0066] Data acquisition module, used to collect pulse voltage data of single lithium-ion battery throughout its life cycle;

[0067] A data processing module, used for preprocessing the pulse voltage data;

[0068] A data extraction module, used to extract a battery voltage rebound characteristic curve based on the pre-processed pulse voltage data;

[0069] A status evaluation module is used to evaluate the current battery health status according to the extracted battery voltage rebound characteristic curve and output the evaluation result;

[0070] The life prediction module is used to predict the battery life according to the current battery health status assessment result and output the prediction result.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] The present invention proposes a battery health status prediction method and system based on the DC internal resistance of a lithium-ion battery. By preprocessing the collected pulse voltage data, the stability and reliability of the data can be effectively improved. According to the preprocessed pulse voltage data, the battery voltage rebound characteristic curve is extracted, that is, a pulse current is applied to collect the voltage rebound data, which can greatly simplify the experimental process, perform tests in real time, and reduce the instrument cost and measurement cost. According to the extracted battery voltage rebound characteristic curve, the battery health status is evaluated, which can effectively simplify the battery health status evaluation process, meet the requirements of being widely used in actual situations, and can effectively predict the battery life according to the current battery health status evaluation result, thereby improving the accuracy and efficiency of the battery health status prediction, and then effectively feedback the internal aging and safety conditions of the battery so as to replace and handle abnormal batteries in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A flowchart of the steps of a method for predicting the battery health state based on the DC internal resistance of a lithium-ion battery provided in an embodiment of the present application;

[0074] Figure 2 A schematic diagram of the states of pulse charging segments collected during battery cycling in the method for lithium battery health diagnosis based on the DC internal resistance of a lithium-ion battery provided in an embodiment of the present application;

[0075] Figure 3 A flow chart of lithium battery health status assessment based on the DC internal resistance of a lithium-ion battery provided in an embodiment of the present application;

[0076] Figure 4 A graph showing the SOH estimation result of a lithium battery based on the DC internal resistance of a lithium-ion battery provided in an embodiment of the present application;

[0077] Figure 5 A lithium battery capacity degradation trajectory and RUL prediction flow chart based on early capacity data of lithium-ion batteries provided in an embodiment of the present application;

[0078] Figure 6 A lithium battery capacity degradation trajectory and RUL prediction result diagram based on early capacity data of lithium-ion batteries provided in an embodiment of the present application;

[0079] Figure 7 A schematic diagram of the structure of a battery health status prediction system based on the DC internal resistance of a lithium-ion battery provided in an embodiment of the present application. DETAILED DESCRIPTION

[0080] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Preferred embodiments of the present invention are provided in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0081] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0082] Embodiment 1:

[0083] A battery health status prediction method based on the DC internal resistance of lithium-ion batteries, see Figure 1 , the method comprises the following steps:

[0084] S1: collecting pulse voltage data of a single lithium-ion battery throughout its life cycle, and preprocessing the pulse voltage data;

[0085] S2: extracting the battery voltage rebound characteristic curve according to the pre-processed pulse voltage data;

[0086] S3: Evaluate the current battery health status based on the extracted battery voltage rebound characteristic curve and output the evaluation result

[0087] S4: Predict the battery life according to the current battery health status assessment result, and output the prediction result.

[0088] In step S1, pulse voltage data of a single lithium-ion battery throughout its life cycle is collected, specifically:

[0089] For batteries with full life cycle, such as Figure 2In the single cycle shown, there are four different states of charge (SOC), and periodic positive and negative pulse signals are applied to control the flow of charging current, and the pulse voltage data of the lithium-ion battery in the whole life cycle is obtained. The battery capacity data is obtained through the battery cabinet's own algorithm according to the ampere-hour integration method, and uploaded to the battery health database. Among them, the pulse voltage data reflects the DC internal resistance of the battery, so the present invention is actually based on the DC internal resistance. The collection of full life cycle data is to charge and discharge the new lithium-ion battery with constant current and constant voltage until the battery capacity drops to 80% of the rated capacity. During this period, the current of constant current and constant voltage charging and constant current discharge is set to any multiple. A pulse charging process is to apply a positive pulse to the battery for a certain time, let it stand for a certain time, and then apply a negative pulse for a certain time, and let it stand for a certain time (wherein the certain time can be any value within 2 to 600 seconds). The current size of the positive pulse and the negative pulse is greater than the charging current and discharge current set for the full life cycle cycle. The acquisition nodes of pulse charging are mainly divided into four charging and discharging states, namely, charging start, standing after full charging, discharging start, and standing after full discharge. The whole experiment was carried out at three temperatures, namely 10°C, 25°C and 45°C, and the temperature was controlled by a thermostat.

[0090] like Figure 2 As shown, this embodiment uses the data of state 3 for illustration, and the data usage methods of other states are consistent with state 3, and the specific prediction results will be slightly different. Because the DC internal resistance is reflected in the pulse voltage, the present invention does not use the traditional resistance voltage ratio definition method, but only uses the original pulse voltage data, including the pulse voltage of the battery at 10°C, 25°C and 45°C.

[0091] In this embodiment, there are 7 batteries at 45°C, numbered 01-07; there are 9 batteries at 25°C, numbered 08-16; and there are 8 batteries at 10°C, numbered 17-24. Here, both the data of the pulse charging stage and the data of the pulse discharging stage can be used. In this step, in order to obtain a more accurate prediction result, the voltage data of the discharge stage with a more stable electrochemical state is collected and used.

[0092] Furthermore, in step S1, the process of preprocessing the pulse voltage data includes:

[0093] The collected pulse voltage data is uploaded to the lithium battery health database, and the historical data in the battery health database is cleaned.

[0094] Specifically, in step S11, the process of performing data cleaning on the historical data in the battery health database includes:

[0095] The historical data in the battery health database is sequentially sorted into tables, missing values ​​are filled, rows and columns are split, duplicate data are deleted, data is normalized, and data is integrated to obtain the voltage data and capacity data during the battery pulse discharge;

[0096] The expression for normalizing the historical data in the battery health database is:

[0097]

[0098] Among them, x i Represents unnormalized voltage data; x max and x min Indicates the maximum and minimum voltage values ​​in different cycles at the same time.

[0099] In some embodiments, the obtained voltage data and capacity data are preprocessed, including removing missing values, duplicate values, and non-numeric data, and processing abnormal noise in the data by polynomial interpolation to ensure the stability and reliability of the data. In order to detect data abnormalities more intuitively, an image of the original data is drawn based on the preprocessed battery data, including an image of the change in pulse voltage and an image of the attenuation of battery capacity, with the horizontal axes being the excitation time and the number of cycles, respectively. Observe the changes in characteristic data during battery aging. After confirming that the data has no obvious abnormalities, use the MIN-MAX method to normalize it to satisfy the formula:

[0100]

[0101] Among them, x i is the unnormalized voltage data, x max and x min It is the maximum and minimum voltage in different cycles at the same time. Normalizing the data helps the model learn the essential characteristics of battery aging, avoids errors caused by large data differences, and speeds up convergence.

[0102] It can be understood that arranging the collected pulse voltage data into tables, filling in missing values, splitting rows and columns, deleting duplicate data, normalizing the data and integrating the data can effectively improve the stability and reliability of the data.

[0103] In step S2, see Figure 3 ,The process of extracting the battery voltage rebound characteristic curve and battery capacity data includes:

[0104] For batteries that cycle throughout their life cycle, a complete pulse charging signal is applied in each cycle to extract the battery voltage rebound characteristic curve during the entire pulse charging process;

[0105] Perform parameter identification on the voltage rebound characteristic curve during a pulse charging process to obtain DC internal resistance data, and divide the DC internal resistance data into a training set and a test set;

[0106] The battery capacity data is calculated using the ampere-hour integration method and the battery cabinet's own algorithm.

[0107] Specifically, in step S12, the complete battery data set (i.e., the DC internal resistance data set) is divided into a training set and a test set. In this embodiment, the leave-one-out method is used for cross-validation, and the data of one battery is used as the test set, and the data of the remaining batteries are used as the training set. First, cross-validation is performed on batteries at the same temperature, and then cross-temperature prediction is performed.

[0108] Specifically, in this embodiment, for 45°C batteries, one battery from the number 01-07 is used as a test battery, and the pulse voltage data of the remaining batteries are merged as a training set, and when merging, the features are ensured to correspond, that is, the same column represents the pulse voltage value at the same time. For example, 01-06 is selected for training, and 07 is selected for testing.

[0109] In this embodiment, for 25°C batteries, one battery from number 08-16 is used as a test battery, and the pulse voltage data of the remaining batteries are merged as a training set. When merging, ensure that the features correspond, that is, the same column represents the pulse voltage value at the same time. For example, 08-15 is selected for training, and 16 is selected for testing.

[0110] In this embodiment, for 10℃ batteries, one battery from number 17-24 is used as a test battery, and the pulse voltage data of the remaining batteries are merged as a training set. When merging, ensure that the features correspond, that is, the same column represents the pulse voltage value at the same time. For example, select 17-23 for training and 24 for testing.

[0111] It can be understood that applying a pulse current and collecting voltage rebound data can greatly simplify the experimental process, conduct tests in real time, and reduce instrument and measurement costs. The DC internal resistance of the battery is reflected by the voltage rebound data, and the experiment can be carried out at any SOC without the need for a complete charge and discharge process, which can effectively simplify the process of battery health status prediction. At the same time, there are differences in the prediction accuracy of different SOCs, and the prediction results of more stable electrochemical states are also relatively stable. Battery health diagnosis based on DC internal resistance can greatly simplify the experimental process, conduct tests in real time, and reduce instrument and measurement costs.

[0112] In step S3, the process of evaluating the battery health status according to the extracted battery voltage rebound characteristic curve includes:

[0113] S31: constructing a battery health status assessment model, wherein the battery health status assessment model includes a convolution module, a bidirectional long short-term memory module and a prediction module;

[0114] S32: inputting the training set into the battery health state assessment model, and setting several rounds of parameter iteration optimization training to optimize the parameters of the battery health state assessment model;

[0115] S33: extracting battery health status features in the training set using a convolution module;

[0116] S34: Using a bidirectional long short-term memory module to extract long-term correlation features during the battery cycle charge and discharge process in the training set;

[0117] S35: The prediction module pre-evaluates the health status of the battery based on the extracted battery health status features and long-term correlation features, and outputs the evaluation results.

[0118] S36: Testing the prediction results using the test set, and setting a loss function to optimize the parameters of the battery health status assessment model after each round of training, and obtaining a trained battery health status assessment model when the preset iterative training round ends or the loss function converges;

[0119] Among them, the battery health status represents the maximum capacity of the battery in the current cycle divided by the initial capacity of the battery, and the evaluation result includes the early capacity data of the battery.

[0120] In some embodiments, the battery health status assessment model constructed in the present application may be a CNN-BiLSTM model.

[0121] Specifically, in step S33, the convolution (CNN) module consists of a convolution layer, a pooling layer, and a fully connected layer. Each convolution layer and pooling layer can be regarded as multiple two-dimensional matrices combined into a tensor. The feature plane realizes the input and output of data and back propagation through unconnected neurons, and neurons share a series of filters to extract data features. The pooling layer is similar to the filter principle, and both enhance the feature extraction effect. Through the deepening of the network structure, the extraction of data topological features gradually shows non-specific features, and finally the data features of the battery health status (i.e., battery health status features) are obtained.

[0122] Specifically, in step S34, in order to better solve the problem of long-term correlation feature extraction in the battery cycle charge and discharge process, a bidirectional long short-term memory (BiLSTM) module is linked after the fully connected layer of the convolution module. The process of extracting long-term correlation features in the battery cycle charge and discharge process in the training set using the bidirectional long short-term memory module includes:

[0123] The bidirectional long short-term memory module includes an input layer, a hidden layer, and an output layer;

[0124] The input layer, hidden layer and output layer are used to extract the long time series in the battery cycle charge and discharge process, and the long-term correlation features are extracted based on the long time series. The expression is:

[0125] f(t)=σ(W f x(t)+W f h(t-1)+b f );

[0126] i(t)=σ(W i x(t)+W i h(t-1)+b i );

[0127] ο(t)=σ(W ο x(t)+W ο h(t-1)+b ο );

[0128] C(t)=f(t)⊙C(t-1)+i(t)⊙tanh(W c x(t)+W c h(t-1)+b c );

[0129] y(t)=ο(t)⊙tanh(C(t));

[0130] The loss function is set to optimize the parameters of the battery health status assessment model after each round of training:

[0131]

[0132] Where x(t) and y(t) represent the input vector and output vector of the current time step t, respectively; f(t) represents the activation value of the forget gate; i(t) represents the activation value of the input gate; o(t) represents the activation value of the output gate; and Y i , and Represent the true value, predicted value and the average of the true value, R 2 represents the coefficient of determination, R 2 represents the coefficient of determination, and RMSE represents the root mean square error.

[0133] Specifically, in this embodiment, the same DC internal resistance data is used as input, and the estimation results in this embodiment are compared with the estimation results of traditional neural network models such as BP, LSTM, GPR, CNN-LSTM, etc., which proves that the present solution is more accurate and stable.

[0134] In this example, parameter identification is performed on experimental data at different temperatures, and a battery health status assessment model is established and trained to prove the feasibility of the method. Finally, in order to achieve cross-temperature assessment, the battery data at different temperatures are mixed together as the input of the neural network, and a battery is selected for testing at different temperatures. For the convenience of illustration, the test set selected is 02, 15, and 21 batteries, corresponding to 45℃, 25℃, and 10℃ respectively. The predicted results are as follows: Figure 4 As shown in the figure, the neural network used is the CNN-BILSTM described above, and the parameters of the network are slightly adjusted. Figure 4 The horizontal axis is the number of cycles, and the vertical axis is SOH. True value represents the true value of SOH; CNN-BiLSTM Predicted represents the SOH value estimated by the CNN-BiLSTM neural network; R 2 represents the coefficient of determination.

[0135] It can be understood that the constructed battery health status assessment model can effectively grasp the changes in pulse voltage data throughout the life cycle of lithium-ion batteries after training the training set, testing the test set, and optimizing the loss function, and then predict the health status of the battery based on the cooperation between the convolution module, the bidirectional long short-term memory module, and the prediction module, effectively simplifying the process of battery health status prediction, and can be widely used in actual situations, thereby effectively improving the accuracy and efficiency of battery health status prediction.

[0136] In this embodiment, the collected pulse voltage data is preprocessed, which can effectively improve the stability and reliability of the data; the battery voltage rebound characteristic curve is extracted based on the preprocessed pulse voltage data, that is, a pulse current is applied to collect voltage rebound data, which can greatly simplify the experimental process, perform tests in real time, and reduce instrument costs and measurement costs; the battery health status is evaluated based on the extracted battery voltage rebound characteristic curve, which can effectively simplify the battery health status evaluation process, meet the needs of wide application in actual situations, and can effectively predict the battery life based on the current battery health status evaluation results, thereby improving the accuracy and efficiency of battery health status prediction, and effectively feedback the internal aging and safety conditions of the battery so as to replace and handle abnormal batteries in time.

[0137] Embodiment 2:

[0138] This embodiment is based on the battery health status prediction method based on the DC internal resistance of a lithium-ion battery described in Embodiment 1, and further describes step S4, as follows:

[0139] Specifically, the pulse voltage rebound data and capacity data of the battery life cycle described in Example 1 are collected. In Example 2, there are 7 batteries at 45°C, labeled 01-07; there are 9 batteries at 25°C, labeled 08-16; and there are 8 batteries at 10°C, labeled 17-24.

[0140] Furthermore, the obtained capacity data is preprocessed, including removing missing values, duplicate values, and non-numeric data, and the abnormal noise in the data is processed by polynomial interpolation to ensure the stability and reliability of the data. In order to more intuitively detect data abnormalities and differences between batteries, the battery capacity curve is drawn with the number of cycles as the horizontal axis and the battery capacity as the vertical axis.

[0141] The complete battery capacity data set (the early battery capacity data estimated in Example 1) is divided into a training set and a test set. In this embodiment, the leave-one-out method is used for cross-validation, and the capacity data of one battery is used as the test set, and the capacity data of the remaining batteries is used as the training set, so that the model learns the decay law of the capacity curve. Considering the close to real usage, the input data of the capacity decay trajectory is the battery capacity value estimated by the CNN-BiLSTM neural network in Example 1, and the future capacity curve of the battery is deduced on this basis.

[0142] Specifically, in the second embodiment, for 45°C batteries, the capacity estimation result of one battery in the number 01-07 is used as the test set, and the capacity data of the remaining batteries are combined as the training set. For example, 01-06 is selected for training, and 07 is selected for testing.

[0143] In the second embodiment, for 25°C batteries, the capacity estimation result of one battery in the number 08-16 is used as the test set, and the capacity data of the remaining batteries are combined as the training set. For example, 08-15 is selected for training, and 16 is selected for testing.

[0144] In the second embodiment, for 10°C batteries, the capacity estimation result of one battery from number 17 to 24 is used as the test set, and the capacity data of the remaining batteries are combined as the training set. For example, 17 to 23 are selected for training, and 24 is selected for testing.

[0145] The process of predicting battery life based on the current battery health status assessment results includes:

[0146] S41: predict the capacity decay trajectory of the battery based on the early capacity data of the battery;

[0147] S42: predicting the battery end-of-life and the current number of cycles according to the battery capacity decay trajectory, and predicting the battery remaining life according to the battery end-of-life and the current number of cycles.

[0148] In step S41, see Figure 5 ,The process of predicting the battery capacity decay trajectory includes:

[0149] S411: constructing a battery capacity decay trajectory prediction model, wherein the battery capacity decay trajectory prediction model includes a self-attention layer, a feedforward neural network layer, and a battery capacity decay trajectory prediction layer;

[0150] S412: inputting the training set into the battery capacity decay trajectory prediction model, and setting several rounds of parameter iterative optimization training to optimize the parameters of the battery capacity decay trajectory prediction model;

[0151] S413: using the self-attention layer to extract the battery capacity features in the training set, the feedforward neural network layer processes the extracted battery capacity features point by point, and the battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity features and outputs the prediction result;

[0152] S414: Use the test set to test the prediction results, and set the loss function to optimize the parameters of the battery capacity decay trajectory prediction model after each round of training. When the preset iterative training round ends or the loss function converges, the trained battery capacity decay trajectory prediction model is obtained.

[0153] In step S413, the self-attention layer uses the self-attention mechanism to extract the battery capacity features in the training set, expressed as:

[0154]

[0155] The feedforward neural network layer processes the output of the self-attention layer point by point, and the expression is:

[0156] FFN(x)=max(0,xW1+b1)W2+b2;

[0157] The battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity characteristics. The expression is:

[0158] Output = softmax(W out DecoderOutput+b out );

[0159] In step S42, the remaining battery life is predicted based on the battery end-of-life and the current number of cycles, and the expression is:

[0160] RUL = battery life - current number of charge and discharge cycles of the battery;

[0161] Where Q is the query matrix, K is the key matrix, and V is the value matrix; d kis the dimension of the key vector, x represents the input feature, W1 and W2 represent the weight matrices of the feedforward neural network, b1 and b2 represent the bias vectors of the feedforward neural network, and RUL is the remaining battery life finally calculated.

[0162] Specifically, in some embodiments, the constructed battery capacity decay trajectory prediction model uses a Transformer model. The Transformer model consists of an encoder and a decoder module. Each encoder and decoder module contains multiple self-attention layers and feedforward neural network layers, which are combined to form a complex network architecture. The self-attention mechanism allows the model to establish global connections between different positions in the input sequence and extract key features in the data by weighted summation. The feedforward neural network layer further processes and transforms the extracted features. Through the stacking of multiple layers of encoders and decoders, the extraction of data features gradually shows high-level and abstract features, and finally the data features of the battery health status are obtained.

[0163] In this embodiment, the self-attention mechanism is implemented by the following formula:

[0164]

[0165] Multi-head self-attention is the parallel calculation result of multiple self-attention mechanisms:

[0166] MultiHead(Q,K,V)=Concat(head1,head 1, ...,head h )W o

[0167] The feedforward neural network layer processes the output of the self-attention layer point by point:

[0168] FFN(x)=max(0,xW1+b1)W2+b2

[0169] The computational formula for the criss-cross attention layer is similar to the encoder’s self-attention layer, but uses the encoder’s output as keys and values:

[0170]

[0171] The positional encoding is added to the input embedding vector using sine and cosine functions to generate the positional encoding:

[0172]

[0173] By stacking multiple layers of encoders and decoders, the model gradually extracts high-level and abstract features, and finally obtains the data features of the battery health status. The final output of the decoder will pass through a linear transformation and a softmax layer to generate a prediction result:

[0174] Output = softmax(W out DecoderOutput+b out )

[0175] In this embodiment, considering the extreme operating capabilities of the detection model, the validation set selects early cycle data to predict future data. Specifically, data from about the first 20% of the battery life cycle is selected. Since the inherent capacity degradation patterns of batteries at different temperatures are quite different, the learning basis of the model is the degradation law of the capacity curve at the macro level. Therefore, in this embodiment, batteries of different temperatures are grouped instead of mixed together.

[0176] In this embodiment, in order to detect the extreme capabilities of the model, verified early battery capacity data is selected for deduction. Specifically, the capacity data of the first 45 cycles of the 10°C battery selection predicts the capacity of the next 200 cycles, the capacity data of the first 50 cycles of the 25°C battery selection predicts the capacity of the next 400 cycles, and the capacity data of the first 100 cycles of the 45°C battery selection predicts the capacity of the next 350 cycles. This embodiment only provides an optional battery early cycle selection scheme, and the battery early cycle selection scheme of the present invention will not be limited to this.

[0177] According to the above steps, the Transformer model is used to predict the future capacity data of the battery at different temperatures, and the capacity degradation curve of the battery is drawn accordingly. In this embodiment, the capacity degradation curve is a composite of two parts: the early capacity data estimated in Example 1 and the future capacity data predicted in this embodiment. According to the total capacity degradation curve, the maximum number of cycles with a capacity value of 75% of the initial value is found as the end life of the battery, and the maximum life of the battery (N EOL ) is used to predict the remaining useful life (RUL) of a battery cycle. For a specific cycle, based on its DCIR data, the model in Example 1 is used to obtain an estimated value of the capacity, which is mapped to the capacity degradation curve predicted in Example 2 to obtain the number of cycles corresponding to the cycle as the "current number of cycles". Based on the maximum life of this battery (N EOL ) minus the current number of cycles (N Current ), and obtain the predicted value of the remaining battery life. The prediction results of this solution are as follows Figure 6 shown. Figure 6The horizontal axis is the number of cycles, and the vertical axis is the battery capacity value. True Data represents the true value of the capacity; Prediction represents the early capacity data estimated in Example 1; Extrapolation represents the predicted battery degradation trajectory; Threshold is the threshold line where the battery capacity drops to 75% of the initial value; Uncertainty represents the uncertainty of the prediction; R 2 , RMSE, and EoL_error represent the coefficient of determination, root mean square error, and end-of-life error, respectively.

[0178] In this embodiment, the constructed battery capacity decay trajectory prediction model can effectively grasp the changes in pulse voltage data of the lithium-ion battery throughout its life cycle after training with a training set, testing with a test set, and optimizing the loss function. Then, based on the cooperation between the self-attention layer, the feedforward neural network layer, and the battery capacity decay trajectory prediction layer, the capacity and life of the battery are predicted, thereby effectively simplifying the process of battery health status prediction and being able to be widely used in actual situations, thereby effectively improving the accuracy and efficiency of battery health status prediction.

[0179] Embodiment three:

[0180] A battery health status prediction system based on the DC internal resistance of lithium-ion batteries, see Figure 7 , the system comprising:

[0181] Data acquisition module, used to collect pulse voltage data of single lithium-ion battery throughout its life cycle;

[0182] A data processing module, used for preprocessing the pulse voltage data;

[0183] A data extraction module, used to extract a battery voltage rebound characteristic curve based on the pre-processed pulse voltage data;

[0184] A status evaluation module is used to evaluate the current battery health status according to the extracted battery voltage rebound characteristic curve and output the evaluation result;

[0185] The life prediction module is used to predict the battery life according to the current battery health status assessment result and output the prediction result.

[0186] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery, characterized in that: The method comprises the following steps: Collecting pulse voltage data of a single lithium-ion battery throughout its life cycle and preprocessing the pulse voltage data; Extracting the battery voltage rebound characteristic curve according to the pre-processed pulse voltage data; According to the extracted battery voltage rebound characteristic curve, the current battery health status is evaluated and the evaluation result is output; Predict the battery life based on the current battery health status assessment result and output the prediction result.

2. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 1, characterized in that: The process of preprocessing the pulse voltage data includes: The collected pulse voltage data is uploaded to the lithium battery health database, and the historical data in the battery health database is cleaned.

3. The battery health status prediction method based on the DC internal resistance of a lithium-ion battery according to claim 2 is characterized in that: The process of data cleaning for historical data in the battery health database includes: The historical data in the battery health database is sequentially sorted into tables, missing values ​​are filled, rows and columns are split, duplicate data are deleted, data is normalized, and data is integrated to obtain the voltage data and capacity data during the battery pulse discharge; The expression for normalizing the historical data in the battery health database is: Among them, x i Represents unnormalized voltage data; x max and x min Indicates the maximum and minimum voltage values ​​in different cycles at the same time.

4. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 1, characterized in that: The process of extracting the battery voltage rebound characteristic curve includes: For batteries that cycle throughout their life cycle, a complete pulse charging signal is applied in each cycle to extract the battery voltage rebound characteristic curve during the entire pulse charging process; Perform parameter identification on the voltage rebound characteristic curve during a pulse charging process to obtain DC internal resistance data, and divide the DC internal resistance data into a training set and a test set; The battery capacity data is calculated using the battery cabinet's own algorithm using the ampere-hour integration method.

5. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 4, characterized in that: Based on the extracted battery voltage rebound characteristic curve, the process of evaluating the current battery health status includes: Constructing a battery health status assessment model, wherein the battery health status assessment model includes a convolution module, a bidirectional long short-term memory module and an assessment module; Inputting the training set into the battery health state assessment model, and setting several rounds of parameter iterative optimization training to optimize the parameters of the battery health state assessment model; Use the convolution module to extract the battery health status features in the training set; The bidirectional long short-term memory module is used to extract the long-term correlation features in the battery pulse charging and discharging process in the training set; The prediction module evaluates the battery health status based on the extracted battery health status features and long-term correlation features, and outputs the evaluation results. The prediction results are tested using the test set, and the loss function is set to optimize the parameters of the battery health status assessment model after each round of training. When the preset iterative training round ends or the loss function converges, the trained battery health status assessment model is obtained; Among them, the battery health status represents the maximum capacity of the battery's current cycle divided by the battery's initial capacity, and the evaluation result includes the battery's early capacity data.

6. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 5, characterized in that: The process of extracting long-term correlation features in the battery pulse charging and discharging process in the training set using the bidirectional long short-term memory module includes: The bidirectional long short-term memory module includes an input layer, a hidden layer, and an output layer; The input layer, hidden layer and output layer are used to extract the long time series in the battery cycle charge and discharge process, and the long-term correlation features are extracted based on the long time series. The expression is: f(t)=σ(W f x(t)+W f h(t-1)+b f ); i(t)=σ(W i x(t)+W i h(t-1)+b i ); o(t)=σ(W o x(t)+W o h(t-1)+b o ); C(t)=f(t)⊙C(t-1)+i(t)⊙tanh(W c x(t)+W c h(t-1)+b c ); y(t)=o(t)⊙tanh(C(t)); The loss function is set to optimize the parameters of the battery health status assessment model after each round of training: Where x(t) and y(t) represent the input vector and output vector of the current time step t, respectively; f(t) represents the activation value of the forget gate; i(t) represents the activation value of the input gate; o(t) represents the activation value of the output gate; and Y i , and Represent the true value, predicted value and the average of the true value, R 2 represents the coefficient of determination, and RMSE represents the root mean square error.

7. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to any one of claim 5, characterized in that: The process of predicting battery life based on the current battery health status assessment results includes: Predict the battery capacity decay trajectory based on the battery's early capacity data; Based on the battery capacity decay trajectory, the battery end-of-life and current cycle number are predicted, and the battery remaining life is predicted based on the battery end-of-life and current cycle number.

8. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 7, characterized in that: The process of predicting the battery capacity decay trajectory includes: Constructing a battery capacity decay trajectory prediction model, wherein the battery capacity decay trajectory prediction model includes a self-attention layer, a feedforward neural network layer, and a battery capacity decay trajectory prediction layer; Inputting the training set into the battery capacity decay trajectory prediction model, and setting several rounds of parameter iterative optimization training to optimize the parameters of the battery capacity decay trajectory prediction model; The self-attention layer is used to extract the battery capacity features in the training set. The feedforward neural network layer processes the extracted battery capacity features point by point. The battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity features and outputs the prediction results. The prediction results are tested using the test set, and the loss function is set to optimize the parameters of the battery capacity decay trajectory prediction model after each round of training. When the preset iterative training round ends or the loss function converges, the trained battery capacity decay trajectory prediction model is obtained.

9. The method for predicting the battery health status based on the DC internal resistance of a lithium-ion battery according to claim 8, characterized in that: The self-attention layer uses the self-attention mechanism to extract the battery capacity features in the training set. The expression is: The feedforward neural network layer processes the output of the self-attention layer point by point, and the expression is: FFN(x)=max(0,xW1+b1)W2+b2; The battery capacity decay trajectory prediction layer predicts the battery capacity decay trajectory based on the point-by-point processed battery capacity characteristics. The expression is: Output=softmax(W out ·DecoderOutput+b out ); The remaining battery life is predicted based on the battery end-of-life and the current number of cycles. The expression is: RUL = battery life - current number of charge and discharge cycles of the battery; Where Q is the query matrix, K is the key matrix, and V is the value matrix; d k is the dimension of the key vector, x represents the input feature, W1 and W2 represent the weight matrices of the feedforward neural network, b1 and b2 represent the bias vectors of the feedforward neural network, and RUL is the remaining battery life finally calculated.

10. A battery health status prediction system based on the DC internal resistance of a lithium-ion battery, characterized in that: The system comprises: Data acquisition module, used to collect pulse voltage data of single lithium-ion battery throughout its life cycle; A data processing module, used for preprocessing the pulse voltage data; A data extraction module, used to extract a battery voltage rebound characteristic curve based on the pre-processed pulse voltage data; A status evaluation module is used to evaluate the current battery health status according to the extracted battery voltage rebound characteristic curve and output the evaluation result; The life prediction module is used to predict the battery life according to the current battery health status assessment result and output the prediction result.

Images