Lithium battery SOH intelligent prediction method and system based on charging curve data

By extracting the voltage and current change rate from the charging curve data of the lithium battery, training the multi-head attention mechanism model, the difficulty in predicting health status caused by incomplete charging cycle data in the microgrid system is solved, and efficient prediction of the health status of lithium battery is achieved .

CN119064793AInactive Publication Date: 2024-12-03NANTONG JINGXIANG TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411478493.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-22
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In microgrid battery energy storage systems, batteries frequently perform charging and discharging operations, resulting in incomplete charging cycle data, existing health factors cannot be effectively used for health status (SOH) prediction, and artificial intelligence-based methods rely on battery attenuation curve Or fully charged and discharge data, requiring a large amount of data and computing resources.

Method used

Using the SOH intelligent prediction method of lithium battery based on charging curve data, the voltage change rate in constant current charging mode and the current change rate in constant voltage charging mode are extracted from the historical data of lithium battery, and after normalization, it is taken as a multi-header. The input of the force mechanism model is trained and tested. When the prediction accuracy reaches the preset value, it is used for real-time health status prediction.

Benefits of technology

In the case of incomplete charging cycle data, the real-time health status value of the lithium battery can be quickly calculated, avoiding a large amount of data calculation and processing, saving computing resources, and improving prediction accuracy and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064793A_ABST
    Figure CN119064793A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of lithium battery intelligent prediction, and provides a lithium battery SOH intelligent prediction method based on charging curve data, and the method comprises the steps: extracting a voltage change rate in a constant-current charging mode and a current change rate in a constant-voltage charging mode from the historical charging data of a lithium battery; the method comprises the following steps: respectively carrying out normalization processing and splicing to form the input of a multi-head attention mechanism model, taking the SOH of the lithium battery as the output of the multi-head attention mechanism model, training and testing the model, and when the prediction precision of the model reaches a preset value, outputting the model for predicting the health state value of the lithium battery. Data in the charging process of the lithium battery are collected in real time, and the SOH value of the lithium battery is predicted. The invention also discloses a system, and the method and the system can complete the precise prediction of the SOH of the lithium battery under the condition that the charge-discharge cycle data of the lithium battery is incomplete. The model does not need a large amount of input data, so that computing resources are saved and the prediction precision is high.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery intelligent prediction, and in particular to a lithium battery SOH intelligent prediction method and system based on charging curve data. Background Art

[0002] Lithium battery intelligent prediction uses advanced machine learning, data analysis and model prediction technology to predict and evaluate the performance, life, health status and remaining power of lithium batteries. It can effectively improve the intelligence level of battery management systems, help users optimize battery use, extend battery life, and ultimately achieve more efficient and safe battery applications. It is crucial for electric vehicles, renewable energy storage systems and other applications that rely on lithium batteries.

[0003] CN118604631A discloses a method for predicting the health status of lithium batteries based on indirect health factors. Step 1: Extract five health factors from the lithium battery data set, including HI1, HI2, HI3, HI4 and HI5, where HI1 is the discharge time of the initial discharge stage of the lithium battery, HI2 is the discharge time of the stable discharge stage of the lithium battery, HI3 is the discharge time of the rapid discharge stage of the lithium battery, HI4 is the actual discharge voltage of the lithium battery after the self-charging ends, and HI5 is the constant current charging time of the lithium battery; Step 2: Perform Pearson correlation coefficient analysis on each health factor and the capacity of the lithium battery, and the calculation formula is: In the formula, x represents the health factor, i = 1, 2, 3, 4, 5; y represents the capacity of the lithium battery; then select the health factors with the top three Pearson correlation coefficient values ​​as the health factor data for the SOH prediction of the lithium battery; Step 3: Remove outliers and smooth the selected health factor data and capacity labels to obtain the preprocessed data health factor and capacity label data; Step 4: The preprocessed health factors are input into the SCINet network for training, and the capacity is used as the prediction label to train the deep learning network to obtain a complete SOH prediction model; Step 5: The five health factors of the predicted lithium battery are input into the trained SOH prediction model, and the output of the SOH prediction model is the SOH prediction result.

[0004] CN114814592B discloses a method and device for estimating the health status and predicting the remaining service life of a lithium battery, including: S1: extracting health features related to capacity degradation on voltage, current and temperature curves during the battery charging cycle; S2: using the grey correlation analysis method to verify the degree of correlation between the extracted health features and battery capacity degradation, and proposing a multi-health feature fusion method to obtain indirect health features; S3: using an improved gravitational search algorithm to optimize a support vector regression model for estimating the health status of a lithium battery; S4: optimizing the support vector regression model based on a polynomial regression model and an improved gravitational search algorithm to estimate the remaining service life of a lithium battery Prediction; In the step S1, the extracted health features include: step S101: in the battery aging process, the capacity directly indicates the degree of battery aging, and the health feature parameters related to capacity degradation are extracted from the voltage, current, and temperature curves of the battery charging process; step S102: extracting the constant current charging time as HF1; extracting the equal voltage difference charging time in the constant current charging stage with a voltage range of 3.8V to 4.2V as HF2; extracting the temperature change rate in the constant current charging stage as HF3; in the step S2, the grey correlation analysis method and the multi-health feature fusion method specifically include: step S201: treating the battery capacity degradation as a reference sequence, The health features are regarded as comparison sequences, step S202: a multi-health feature fusion method is proposed to obtain indirect health features IHF; in the step S3, the improved gravitational search algorithm optimizes the support vector regression model specifically including: step S301: establishing a support vector regression model, selecting radial basis function as the kernel, step S302: using the improved gravitational search algorithm to optimize the penalty factor and kernel function parameters of the support vector regression, in order to avoid the traditional gravitational search algorithm easily falling into the local optimal solution and insufficient memory problems, the improved algorithm is proposed as follows: 1) the chaotic sequence generates an initial population; 2) global memory is introduced; step S303: the battery data is divided into a training set and a test set, and the support vector regression model is optimized by the improved gravitational search algorithm for lithium battery Battery health status estimation; in the step S4, the polynomial regression model specifically includes: step S401: when predicting the remaining service life of the lithium battery, indirect health features need to be used as input, but the IHF of the future number of cycles cannot be measured at present, and a polynomial regression model is used to describe the relationship between IHF and the number of cycles, so as to predict the IHF at a future moment, step S402: because there is a certain coupling relationship between the remaining service life and the health status of the lithium battery, and there is a mapping relationship between the health status and IHF, a coupling framework is adopted, and the values ​​of IHF and the current health status are used to optimize the support vector regression model through an improved gravitational search algorithm to predict the remaining service life of the lithium battery, so as to realize the joint estimation of the battery health status and the remaining service life.

[0005] CN118393386A discloses a completely data-driven lithium battery health state prediction method, the method comprising the following steps: S1, obtaining the voltage and current data of the lithium battery during charging as characteristic factors to construct a characteristic vector and generate training samples and test samples; S2, constructing an initial prediction model for the battery health state, and optimizing the initial prediction of the battery health state; S3, training the optimized initial prediction model for the battery health state through the training samples of step S1, obtaining the initial prediction value of the battery health state of the training sample and the trained initial prediction model for the battery health state; S4, obtaining the initial prediction error of the training sample according to the initial prediction model for the battery health state, and converting the training sample into an initial prediction value of the battery health state; S5, obtaining the initial prediction error of the training sample according to the initial prediction model for the battery health state, and optimizing the initial prediction of the training sample. The initial prediction value of the battery health state, the initial prediction error of the training sample, and the characteristic factor constitute a new training set; S5, construct an error compensation model, and optimize the error compensation model; S6, train the optimized extreme learning machine model through the new training set of step S4 to obtain a trained extreme learning machine model; input the test sample of step S2 into the initial prediction model of the battery health state trained in step S3 and the trained error compensation model to obtain the initial prediction value and initial prediction error of the battery health state of the test sample; S7, linearly superimpose the initial prediction value and initial prediction error of the battery health state of the test sample to obtain the final prediction value of the battery health state.

[0006] In microgrid battery energy storage systems, batteries are frequently charged and discharged, often discharging before being fully charged. This state results in incomplete data on charging cycles, making existing health factors ineffective for data-driven state of health (SOH) prediction. Summary of the invention

[0007] Long-term practice has found that the battery's state of health (SOH) directly affects the choice of charging and discharging strategies and the reliability of the entire energy storage system. Data-driven methods based on artificial intelligence to predict the health status of lithium batteries usually rely on battery attenuation curves or ideal full charge and discharge data, and require a large amount of data and computing resources, complex models and other technical issues.

[0008] In view of this, the present invention aims to propose a lithium battery SOH intelligent prediction method based on charging curve data, comprising: Step S1, extracting the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extracting the voltage change rate {dV} under the constant current charging mode with ∆t as the time period; normalizing the voltage change rate {dV} to obtain a first data set {A}; Taking ∆t1 as the time period, extract the current change rate {dI} in the constant voltage charging mode, and obtain the second data set {B} after normalizing the current change rate {dI}; Step S2, concatenating the first data set {A} and the second data set {B} to form a third data set {C}, and encoding the elements in the third data set {C} by position; using the third data set {C} as the input of the multi-head attention mechanism model, and using the health state value {SOH} as the output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set, and the test set of the third data set {C} is set to 8:1:1; Step S3: if the prediction accuracy of the multi-head attention mechanism model reaches a preset value, output the trained multi-head attention mechanism model; Step S4, using a sensor to obtain real-time data of the lithium battery charging process, from the real-time data of the lithium battery charging process, with ∆t as the time period, extract the voltage change rate {dV1} in the constant current charging mode; with ∆t1 as the time period, extract the current change rate {dI1} in the constant voltage charging mode, normalize the voltage change rate {dV1} and the current change rate {dI1} respectively, and splice them to form a third data set {C1} as the input of the trained multi-head attention mechanism model, and calculate the health status prediction value {SOH1}.

[0009] Preferably, in the constant current charging mode, the i-th voltage change rate is,

[0010] Wherein, V(t) is the function of the voltage value changing with time t in the constant current charging mode; The jth current change rate in constant voltage charging mode is,

[0011] Wherein, I(t) is a function of the current value changing with time t in the constant voltage charging mode.

[0012] Preferably, each charging cycle includes a constant current charging mode and a constant voltage charging mode.

[0013] Preferably, the charging time in the constant voltage charging mode is kept at least t g .

[0014] Preferably, in step S4, sensors are used to simultaneously acquire real-time data of the charging process of multiple lithium batteries.

[0015] Preferably, the normalization process adopts minimum-maximum normalization or Z-score.

[0016] The present invention also discloses a system based on the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data, the system comprising: A data extraction unit is used to extract the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extract the voltage change rate {dV} in the constant current charging mode with ∆t as the time period; the voltage change rate {dV} is normalized to obtain a first data set {A}; Taking ∆t1 as the time period, extract the current change rate {dI} in the constant voltage charging mode, and obtain the second data set {B} after normalizing the current change rate {dI}; A preprocessing unit, configured to concatenate the first data set {A} and the second data set {B} to form a third data set {C}, and position-encode the elements in the third data set {C} as input of a multi-head attention mechanism model, and use the health state value {SOH} as output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set, and the test set of the third data set {C} is set to 8:1:1; A training unit, used for outputting the trained multi-head attention mechanism model if the prediction accuracy of the multi-head attention mechanism model reaches a preset value; The prediction unit is used to obtain the real-time data of the lithium battery charging process using a sensor, and extract the voltage change rate {dV1} in the constant current charging mode with ∆t as the time period from the real-time data of the lithium battery charging process; and extract the current change rate {dI1} in the constant voltage charging mode with ∆t1 as the time period. After normalizing the voltage change rate {dV1} and the current change rate {dI1} respectively, they are spliced ​​to form a third data set {C1} as the input of the trained multi-head attention mechanism model, and the health status prediction value {SOH1} is calculated.

[0017] Preferably, the system further comprises a signal transmission unit for transmitting real-time data of the lithium battery charging process acquired by the sensor to the data extraction unit.

[0018] The present invention provides an electronic device, at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data.

[0019] The present invention provides a machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data of the present application.

[0020] The present invention discloses a lithium battery SOH intelligent prediction method based on charging curve data. From the lithium battery historical charging data, for each charging cycle, the voltage change rate in the constant current charging mode and the current change rate in the constant voltage charging mode, and the health state value of the lithium battery are extracted, respectively, and spliced ​​to form the input of the multi-head attention mechanism model after normalization processing. The health state value of the lithium battery is used as the output of the multi-head attention mechanism model, and the model is trained and tested. When the prediction accuracy of the model reaches a preset value, the output model is used to predict the real-time health state value SOH of the lithium battery. Real-time data in the charging process of the lithium battery is collected by a sensor, and the real-time data is input into the multi-head attention mechanism model after preprocessing, so as to obtain the predicted value of the real-time health state value SOH of the lithium battery. The present invention also discloses a system for the above method. When the lithium battery is frequently charged and discharged, and starts to discharge when it is not fully charged, resulting in incomplete charging cycle data, the method and system extract a finite number of discrete values ​​of the voltage change rate from the constant current charging mode, and extract a finite number of discrete values ​​of the current change rate from the constant voltage charging mode, and splice these data to form the input of the model. Therefore, the lithium battery does not need to be fully charged during constant voltage charging. The predicted value of the real-time health state SOH can be quickly calculated based on the voltage change rate in constant current charging mode and the current change rate in constant voltage charging mode. The model can complete the prediction of the battery health state SOH with only a limited amount of input data, avoiding a large amount of data calculation and processing, thereby saving computing resources, improving efficiency and high prediction accuracy.

[0021] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 A schematic diagram of a lithium battery SOH intelligent prediction method based on charging curve data according to an embodiment of the present invention; Figure 2 A comparison diagram of predicted values ​​and actual values ​​of a lithium battery SOH intelligent prediction method based on charging curve data according to an embodiment of the present invention; Figure 3 1 is the charging curve of the present invention in the constant current charging mode after different cycle times (40, 80, 120, 160 cycles); Figure 4These are the charging curves under the constant voltage charging model after different cycle numbers (40, 80, 120, and 160 cycles) of the present invention. DETAILED DESCRIPTION

[0023] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", "third", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] In order to solve the technical problems existing in the prior art, that is, in the microgrid battery energy storage system, the battery is frequently charged and discharged, and often starts to discharge before it is fully charged. This state results in incomplete data on the charging cycle, making the existing health factors unable to be effectively used for data-driven state of health (SOH) predictions. The state of health (SOH) of the battery directly affects the choice of charging and discharging strategies and the reliability of the entire energy storage system. The data-driven method based on artificial intelligence for predicting the health status of lithium batteries usually relies on battery attenuation curves or ideal complete charge and discharge data, and requires a large amount of data and computing resources, complex models and other technical problems. The present invention provides a lithium battery SOH intelligent prediction method based on charging curve data, such as Figure 1 As shown, a schematic diagram of a lithium battery SOH intelligent prediction method based on charging curve data according to an embodiment of the present invention, the lithium battery SOH intelligent prediction method based on charging curve data includes: Step S1, extracting the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extracting the voltage change rate {dV} under the constant current charging mode with ∆t as the time period; normalizing the voltage change rate {dV} to obtain a first data set {A}; Taking ∆t1 as the time period, extract the current change rate {dI} in the constant voltage charging mode, and obtain the second data set {B} after normalizing the current change rate {dI}; Step S2, concatenating the first data set {A} and the second data set {B} to form a third data set {C}, and encoding the elements in the third data set {C} by position; using the third data set {C} as the input of the multi-head attention mechanism model, and using the health state value {SOH} as the output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set, and the test set of the third data set {C} is set to 8:1:1; Step S3: if the prediction accuracy of the multi-head attention mechanism model reaches a preset value, output the trained multi-head attention mechanism model; Step S4, using a sensor to obtain real-time data of the lithium battery charging process, from the real-time data of the lithium battery charging process, with ∆t as the time period, extract the voltage change rate {dV1} in the constant current charging mode; with ∆t1 as the time period, extract the current change rate {dI1} in the constant voltage charging mode, normalize the voltage change rate {dV1} and the current change rate {dI1} respectively, and splice them to form a third data set {C1} as the input of the trained multi-head attention mechanism model, and calculate the health status prediction value {SOH1}.

[0027] The present invention discloses a lithium battery SOH intelligent prediction method based on charging curve data. From the lithium battery historical charging data, for each charging cycle, the voltage change rate in the constant current charging mode and the current change rate in the constant voltage charging mode, and the health state value of the lithium battery are extracted, respectively, and spliced ​​to form the input of the multi-head attention mechanism model after normalization processing. The health state value of the lithium battery is used as the output of the multi-head attention mechanism model, and the model is trained and tested. When the prediction accuracy of the model reaches a preset value, the output model is used to predict the real-time health state value SOH of the lithium battery. Real-time data in the charging process of the lithium battery is collected by a sensor, and the real-time data is input into the multi-head attention mechanism model after preprocessing, so as to obtain the predicted value of the real-time health state value SOH of the lithium battery. When the lithium battery is frequently charged and discharged, and starts to discharge when it is not fully charged, resulting in incomplete charging cycle data, this method extracts a finite number of discrete values ​​of the voltage change rate from the constant current charging mode, and extracts a finite number of discrete values ​​of the current change rate from the constant voltage charging mode, and splices these data to form the input of the model. Therefore, the lithium battery does not need to be fully charged during constant voltage charging. The predicted value of the real-time health state SOH can be quickly calculated based on the voltage change rate in constant current charging mode and the current change rate in constant voltage charging mode. The model can complete the prediction of the battery health state SOH with only a limited amount of input data, avoiding a large amount of data calculation and processing, thereby saving computing resources, improving efficiency and high prediction accuracy.

[0028] Generally, charging of lithium batteries usually requires two processes, first the constant current charging mode and the constant voltage charging mode. The constant current charging mode (CC - Constant Current) is short in time, and the charger charges the battery with a fixed current (constant current) until the battery voltage reaches the set upper limit. Charge the battery quickly to improve the battery status and capacity as soon as possible. The voltage change rate in the constant current charging mode can accurately reflect the changing trend of the battery voltage during the charging process, which helps to identify the battery status and performance. The constant voltage charging mode (CV - Constant Voltage) is long in time, and the charger maintains a constant voltage, while the current will gradually decrease. Ensure that the battery will not be damaged by overcharging when it is fully charged. Generally speaking, the charging process of lithium batteries includes constant current charging mode and partial constant voltage charging mode, so the charging data in the constant voltage charging mode is not complete. The present invention adopts the constant voltage charging mode to extract a limited number of current change rates, which can provide sufficient health characteristics for predicting the battery health status. In order to extract only limited data during the lithium battery charging process and accurately predict the SOH value, in the more preferred case of the present invention, the i-th voltage change rate in the constant current charging mode is,

[0029] Wherein, V(t) is a function of the voltage value changing with time t in the constant current charging mode; the charging curve in the constant current charging mode is formed by the V(t) function, such as Figure 4 As shown, they are the charging curves under constant current charging mode after different number of cycles (40, 80, 120, and 160 cycles).

[0030] The jth current change rate in constant voltage charging mode is,

[0031] Where I(t) is a function of the current value changing with time t in the constant voltage charging mode. The charging curve in the constant voltage charging mode is formed by the I(t) function. Figure 4 As shown, I(t) after different number of cycles (40, 80, 120, and 160 cycles) is the charging curve under constant voltage charging mode.

[0032] In order to better balance the efficiency of obtaining the voltage change rate and the current change rate and the accuracy of lithium battery SOH prediction, in a more preferred case of the present invention, 5-10 voltage change rate values ​​are extracted under the constant current charging mode, and 5-10 current change rate values ​​are extracted under the constant voltage charging mode.

[0033] In order to better optimize the charging process, the constant current and constant voltage charging modes are combined, and the charger monitors the current and voltage in real time to detect any abnormal conditions in a timely manner. In a more preferred embodiment of the present invention, each charging cycle includes a constant current charging mode and a constant voltage charging mode. For example, the charging time is short in the constant current charging mode stage, and it can generally be completed within 5-10 minutes. The charging of this stage will be completed during the charging and discharging process of the lithium battery. With ∆t=1min as the sampling period, 5-10 voltage change rate values ​​are collected. In the constant voltage charging mode, the charger will maintain the output voltage at 4.2V. When the battery is close to full charge, the internal resistance increases, and the charging current will gradually decrease, eventually approaching zero. This process takes a long time and accounts for a large proportion of the total charging time.

[0034] In order to obtain sufficient current change rate values ​​in the constant voltage charging mode, in a more preferred embodiment of the present invention, the charging time in the constant voltage charging mode is kept at least t g For example, when the constant current charging mode ends, the charging time in the constant voltage charging mode is maintained for at least t g =5~10min, then ∆t1=t g / N, N is the number of sampled current change rate values. That is, for a lithium-ion battery (such as a battery with LiCoO2 chemical composition), keep the maximum charging voltage at 4.2V and charge for 5 to 10 minutes. If the charge is 5 minutes and the number of sampled current change rate values ​​is 10, then sampling is performed every 0.5 minutes to obtain 10 values. , thus forming {dI}.

[0035] In order to simultaneously predict the values ​​of multiple batteries in a lithium battery energy storage system, in a more preferred embodiment of the present invention, in step S4, a sensor is used to simultaneously obtain real-time data of the charging process of multiple lithium batteries. In the multi-head attention mechanism model (Multi-Head Attention), the input sequence first passes through three different linear transformation layers to obtain Query, Key, and Value respectively. Then, these transformed vectors are divided into several "heads", each with its own independent Query, Key, and Value matrices. For each head, a scaled dot-product attention (Scaled Dot-ProductAttention) operation is performed. The outputs of all heads are concatenated together and then fused through a linear layer to obtain the final attention output vector. Figure 2 This is the prediction of SOH using the multi-head attention mechanism model. Estimated is the predicted value, the Real orange line is the actual value, and the prediction accuracy MSE is 0.324×10 -4 , the accuracy is much better than BP, CNN, and LSTM models.

[0036] In order to convert data of different ranges or distributions into a unified standard, and in order to better splice the first data set {A} and the second data set {B} to form a third data set {C}. In a more preferred case of the present invention, the normalization process adopts minimum-maximum normalization or Z-score. In order to further reduce the calculation amount of the model and save computing resources, in a more preferred case of the present invention, the third data set {C} is subjected to principal component analysis (PCA), and the high-dimensional data of the third data set {C} is projected into a lower-dimensional space to obtain the principal component {PC}, and the elements in the set {PC} are positionally encoded and used as the input of the multi-head attention mechanism model.

[0037] The present invention also discloses a system based on the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data, the system comprising: A data extraction unit is used to extract the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extract the voltage change rate {dV} in the constant current charging mode with ∆t as the time period; the voltage change rate {dV} is normalized to obtain a first data set {A}; Taking ∆t1 as the time period, extract the current change rate {dI} in the constant voltage charging mode, and obtain the second data set {B} after normalizing the current change rate {dI}; A preprocessing unit, configured to concatenate the first data set {A} and the second data set {B} to form a third data set {C}, and position-encode the elements in the third data set {C} as input of a multi-head attention mechanism model, and use the health state value {SOH} as output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set, and the test set of the third data set {C} is set to 8:1:1; A training unit, used for outputting the trained multi-head attention mechanism model if the prediction accuracy of the multi-head attention mechanism model reaches a preset value; The prediction unit is used to obtain the real-time data of the lithium battery charging process using a sensor, and extract the voltage change rate {dV1} in the constant current charging mode with ∆t as the time period from the real-time data of the lithium battery charging process; and extract the current change rate {dI1} in the constant voltage charging mode with ∆t1 as the time period. After normalizing the voltage change rate {dV1} and the current change rate {dI1} respectively, they are spliced ​​to form a third data set {C1} as the input of the trained multi-head attention mechanism model, and the health status prediction value {SOH1} is calculated.

[0038] The system for the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data disclosed in the present invention, the data extraction unit extracts the voltage change rate in the constant current charging mode and the current change rate in the constant voltage charging mode, and the health state value of the lithium battery from the historical charging data of the lithium battery for each charging cycle, and after normalization processing, they are spliced ​​in the preprocessing unit to form the input of the multi-head attention mechanism model, and the health state value of the lithium battery is used as the output of the multi-head attention mechanism model to train and test the model. After the prediction accuracy of the model in the training unit reaches the preset value, the output model is used to predict the real-time health state value SOH of the lithium battery. The prediction unit collects real-time data during the charging process of the lithium battery through a sensor, and inputs the real-time data into the multi-head attention mechanism model after preprocessing, so as to obtain the predicted value of the real-time health state value SOH of the lithium battery. When the lithium battery is frequently charged and discharged, and starts to discharge when it is not fully charged, resulting in incomplete charging cycle data, the system extracts a finite number of discrete values ​​of the voltage change rate from the constant current charging mode, and extracts a finite number of discrete values ​​of the current change rate from the constant voltage charging mode, and splices these data to form the input of the model. Therefore, the lithium battery does not need to be fully charged during constant voltage charging. The predicted value of the real-time health state SOH can be quickly calculated based on the voltage change rate in constant current charging mode and the current change rate in constant voltage charging mode. The model can complete the prediction of the battery health state SOH with only a limited amount of input data, avoiding a large amount of data calculation and processing, thereby saving computing resources, improving efficiency and high prediction accuracy.

[0039] In order to quickly transmit the real-time data (such as voltage, current, temperature, etc.) acquired by the sensor to the data extraction unit, in a preferred embodiment of the present invention, the system further comprises a signal transmission unit for transmitting the real-time data of the lithium battery charging process acquired by the sensor to the data extraction unit.

[0040] The present invention provides an electronic device, at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data.

[0041] The present invention provides a machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the above-mentioned lithium battery SOH intelligent prediction method based on charging curve data of the present application.

[0042] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0043] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0044] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent prediction method for lithium battery SOH based on charging curve data, It is characterized in that The lithium battery SOH intelligent prediction method based on charging curve data includes: Step S1, extracting the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extracting the voltage change rate {dV} under the constant current charging mode with ∆t as the time period; normalizing the voltage change rate {dV} to obtain a first data set {A}; With ∆t 1 is a time period, extracting the current change rate {dI} in the constant voltage charging mode, and normalizing the current change rate {dI} to obtain a second data set {B}; Step S2, concatenating the first data set {A} and the second data set {B} to form a third data set {C}, and encoding the elements in the third data set {C} by position; using the third data set {C} as the input of the multi-head attention mechanism model, and using the health state value {SOH} as the output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set, and the test set of the third data set {C} is set to 8:1:1; Step S3: if the prediction accuracy of the multi-head attention mechanism model reaches a preset value, output the trained multi-head attention mechanism model; Step S4, using a sensor to obtain real-time data of the lithium battery charging process, and extracting the voltage change rate {dV 1 }; with ∆t 1 is the time period, extract the current change rate {dI 1 }, the voltage change rate {dV 1 } and the current change rate {dI 1 }After normalization, they are concatenated to form the third data set {C 1 } is used as the input of the trained multi-head attention mechanism model to calculate the health status prediction value {SOH 1 }.

2. The lithium battery SOH intelligent prediction method based on charging curve data according to claim 1, It is characterized in that The i-th voltage change rate in constant current charging mode is, Wherein, V(t) is the function of the voltage value changing with time t in the constant current charging mode; The jth current change rate in constant voltage charging mode is, Wherein, I(t) is a function of the current value changing with time t in the constant voltage charging mode.

3. The lithium battery SOH intelligent prediction method based on charging curve data according to claim 2, It is characterized in that Each charging cycle includes constant current charging mode and constant voltage charging mode.

4. The lithium battery SOH intelligent prediction method based on charging curve data according to claim 1, It is characterized in that The charging time in constant voltage charging mode should be at least t g .

5. The lithium battery SOH intelligent prediction method based on charging curve data according to claim 1, It is characterized in that In step S4, sensors are used to simultaneously acquire real-time data of the charging process of multiple lithium batteries.

6. The lithium battery SOH intelligent prediction method based on charging curve data according to any one of claims 1 to 5, It is characterized in that The normalization process adopts minimum-maximum normalization or Z-score.

7. A system based on the lithium battery SOH intelligent prediction method based on charging curve data according to any one of claims 1 to 6, It is characterized in that The system comprises, A data extraction unit is used to extract the health status value {SOH} of the lithium battery for each charging cycle from the historical charging data of the lithium battery, and extract the voltage change rate {dV} in the constant current charging mode with ∆t as the time period; the voltage change rate {dV} is normalized to obtain a first data set {A}; With ∆t 1 is a time period, extracting the current change rate {dI} in the constant voltage charging mode, and normalizing the current change rate {dI} to obtain a second data set {B}; A preprocessing unit, configured to concatenate the first data set {A} and the second data set {B} to form a third data set {C}, and position-code the elements in the third data set {C}; As the input of the multi-head attention mechanism model, the health state value {SOH} is used as the output to train the multi-head attention mechanism model; wherein the ratio of the training set, the validation set and the test set of the third data set {C} is set to 8:1:1; A training unit, used for outputting the trained multi-head attention mechanism model if the prediction accuracy of the multi-head attention mechanism model reaches a preset value; The prediction unit is used to obtain the real-time data of the lithium battery charging process using a sensor, and extract the voltage change rate {dV 1 }; with ∆t 1 is the time period, extract the current change rate {dI 1 }, the voltage change rate {dV 1 } and the current change rate {dI 1 }After normalization, they are concatenated to form the third data set {C 1 } is used as the input of the trained multi-head attention mechanism model to calculate the health status prediction value {SOH 1 }.

8. The system according to claim 7, It is characterized in that The system also includes a signal transmission unit, which is used to transmit the real-time data of the lithium battery charging process acquired by the sensor to the data extraction unit.

9. An electronic device, It is characterized in that at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the lithium battery SOH intelligent prediction method based on charging curve data as described in any one of claims 1-6.

10. A machine-readable storage medium, It is characterized in that The machine-readable storage medium stores instructions for enabling a machine to execute the lithium battery SOH intelligent prediction method based on charging curve data as described in any one of claims 1 to 6 of the present invention.

Citation Information

Cited By

  • Lithium battery health state cloud monitoring method and system

    CN119881713A

  • A lithium battery health status cloud monitoring method and system

    CN119881713B