A lithium battery health state estimation method based on equal pressure rise energy and improved GRU

By extracting the isovoltage energy and peak sequence of lithium batteries based on isovoltage energy increase and improving GRU network, combining the flip layer and probability layer to build a health status model, and solving the data error in the health status estimation of lithium batteries and The problem of insufficient modeling accuracy achieves more accurate health status estimation.

CN115389946BActive Publication Date: 2024-11-19GUANGDONG DING XI TONGXIN IND CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211142117.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-20
Publication Date
2024-11-19
Estimated Expiration
2042-09-20

AI Technical Summary

Technical Problem

The existing lithium battery health status estimation methods have problems such as large data errors and insufficient modeling accuracy, especially when the battery aging and regeneration phenomena are complex, resulting in inaccurate estimation.

Method used

Using the GRU network method based on isovoltage rise energy and improved, by extracting isovoltage rise energy sequence and peak sequence, combining the flip layer and probability layer, a network estimation model for the healthy state of lithium batteries is constructed, and noise error and redundant information are overcome to improve the Estimate accuracy.

Benefits of technology

It effectively overcomes the inaccurate health status estimation problem caused by battery aging and regeneration, improves the generalization and robustness of the model, and provides a more accurate estimate of the health status of lithium batteries.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115389946B_ABST
    Figure CN115389946B_ABST
Patent Text Reader

Abstract

The present invention provides a lithium battery health state estimation method based on isobaric energy and improved GRU, including obtaining health state data and charging energy data of each charge and discharge cycle of the lithium battery; extracting isobaric interval energy sequence from the charging energy data of each charge and discharge cycle, drawing isobaric energy charging curve, and calculating the peak sequence of isobaric energy charging curve corresponding to each charge and discharge cycle; normalizing and preprocessing the peak sequence of isobaric energy charging curve and health state sequence as input data; adding a flip layer and a probability layer to the GRU network, improving the GRU network, and constructing a network estimation model for the health state of the lithium battery; and using the constructed lithium battery health state network estimation model to estimate the health state of the lithium battery. The present invention characterizes the complex and violent electrochemical reactions inside the battery through the external characteristics of the battery, can effectively reflect the decay and aging of the battery, and has the advantages of fast and accurate health state estimation method.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of battery technology, and more specifically, relates to a lithium battery health status estimation method based on equal pressure rise energy and improved GRU. Background Art

[0002] With the improvement of green environmental awareness and the promotion of sustainable development of new energy, lithium-ion batteries have been studied and paid attention to in various fields due to their excellent performance advantages such as high power, high energy density, long service life and low environmental pollution. In the recycling process of lithium batteries, the health status becomes a key indicator to evaluate the degree of battery aging. The aging and regeneration of batteries are complex and changeable, and are closely related to the internal and external conditions of the battery. Therefore, it is very important to accurately estimate the health status of lithium batteries.

[0003] In recent years, data-driven methods have been widely used to estimate the health status of lithium batteries. Because there is no need to analyze the internal electrochemical mechanism of lithium batteries, only the current, voltage and other characteristics collected from the outside of the battery are modeled, which has great advantages in computing efficiency and cost. Commonly used methods include support vector machine (SVM), extreme learning machine (ELM), long short-term memory network (LSTM), etc. However, these modeling methods are limited by the amount of data and the test environment, resulting in the inability of the estimation model to accurately describe the battery degradation process. As a variant of LSTM, GRU uses fewer parameters and reduces the complexity of the network, which greatly improves the data computing burden and convergence speed. However, the regeneration phenomenon and data jump that occur with battery aging can easily lead to inaccurate estimates of the health status.

[0004] Although data-driven methods can efficiently establish estimation models, the analysis and extraction of features largely determine the estimation accuracy of the model. For example, the commonly used incremental analysis method extracts features from charging curves such as voltage, current, and temperature, and reflects the violent reaction inside the battery through subtle changes in external characteristics. However, the traditional incremental analysis method is usually used to analyze the battery aging mechanism under constant current charging and discharging mode. The capacity-terminal voltage curve of the original charging and discharging data is transformed into a capacity increment-terminal voltage curve using a first-order differential derivative. However, in actual measurements, there are noise errors and fluctuation errors caused by differential derivatives of charging curves, as well as accuracy issues that greatly hinder the application of this method. Summary of the invention

[0005] In order to solve the deficiencies in the current technical status, and aiming at the problems of large data acquisition errors and insufficient modeling accuracy in the current health status estimation method, the present invention proposes a lithium battery health status estimation method based on isobaric energy rise and improved GRU. The complex and violent electrochemical reactions occurring inside are characterized by external characteristics, and features are quickly extracted. The estimation model can effectively overcome the regeneration phenomenon of the battery and improve the estimation accuracy of the battery health status.

[0006] In order to solve the above practical problems, the technical solution adopted by the present invention is:

[0007] A method for estimating the health status of a lithium battery based on equal pressure rise energy and improved GRU, the method comprising the following steps:

[0008] S1. Obtain health status data and charging energy data of each charge and discharge cycle of the lithium battery;

[0009] S2. Extracting the equal-pressure rise interval energy sequence from the charging energy data of each charge-discharge cycle, drawing the equal-pressure rise energy charging curve, and calculating the peak sequence of the equal-pressure rise energy charging curve corresponding to each charge-discharge cycle;

[0010] S3, normalizing and preprocessing the peak sequence of the equal pressure rise energy charging curve and the health state sequence as input data;

[0011] S4. Add flipping layer and probability layer to the GRU network, improve the GRU network, and build a network estimation model for the health status of lithium batteries;

[0012] S5. Use the constructed lithium battery health status network estimation model to estimate the health status of the lithium battery.

[0013] Wherein, in step S1:

[0014] The health status data sequence of lithium batteries during the charge and discharge cycle is H1, H2, ..., H n , where the measured data Among them, H i is the battery health status of the i-th charge and discharge cycle, i = 1, 2, K, n, n is the number of charge and discharge cycles, C i is the maximum discharge capacity of the lithium battery in the i-th charge and discharge process, C is the rated capacity of the lithium battery; the charging energy data are E0, E1, ..., E t , where the measured energy data E j =E0+∑UIt, where E j is the charging energy at time j, E0 is the initial energy, U is the charging terminal voltage, I is the charging current, and t is the charging time.

[0015] Preferably, the process of step S2 includes:

[0016] The energy sequence E0,E corresponding to the fixed voltage rise ΔV ,…,E mΔV To quickly draw the constant pressure rise energy charging curve, calculate the peak sequence of the constant pressure rise energy charging curve as F1, F2, ..., F n ,in, F i is the maximum rate of change of the isobaric energy rise within i charging cycles, E i,aΔV 、E i,(a+1)ΔV are the charging energies of the battery in the adjacent ath and a+1th equal-voltage intervals in the ith charging cycle, respectively, a=0,…,m-1, ΔV is the interval voltage, and y is the total rising voltage during the charging process.

[0017] Further, the process of step S3 includes:

[0018] The data is preprocessed by using the maximum and minimum normalization method to normalize the data to the interval [-1,1]. The normalization formula is: x is the preprocessed data sequence, x min is the minimum value in the data sequence x, x max is the maximum value in the data sequence x.

[0019] Furthermore, in step S4:

[0020] The flip layer reverses the order of the input data sequence and reads both forward and reverse data at the same time. The flip calculation formula is as follows:

[0021] l = size(x * );

[0022]

[0023] Among them, l is a constant representing the length of the input sequence, and size is the dimension function. For x * Reverse sequence of the same dimension, flip is the flip function, dim is the dimension direction of the reverse operation;

[0024] The probability layer retains and deletes the input data according to a certain probability. The probability calculation formula is as follows:

[0025] g i = Bernoulli(p);

[0026]

[0027]

[0028] Among them, p is a probability value, 0 <p<1,g i A random sequence of 0,1 with the same length as the data sequence is generated for a Bernoulli distribution with probability p, is the data retained according to probability, y is the output data, w is the weight of the retained data, and b is the bias;

[0029] The main data processing process in the GRU network is as follows:

[0030] z t =σ1(w z *[h t-1 , x t ]+b z )

[0031] r t =σ1(w r *[h t-1 , x t ]+b r )

[0032]

[0033]

[0034] Among them, the update gate z t Measure the degree to which the information at time t-1 is retained at time t, reset gate r t Used to control the information at time t-1 to be retained in the candidate hidden information at time t According to the updated gate state at time t and the candidate hidden state at time t, the hidden state h of the output gate is calculated. t ;w z , w r and w h are the weights of the update gate, reset gate and hidden state, respectively, and b z , b r and b h are the update gate, reset gate and hidden state bias respectively; σ1 and σ2 are sigmoid and tanh activation functions respectively, and ⊙ is the matrix element dot product;

[0035] Take the first k (k=1,…,n-1) sample values ​​of the peak sequence and the health state sequence and As the training set, k=1,…,n-1, take the k+1,k+2,…,nth sample values and As a test set, a battery health state estimation model is constructed through the improved GRU network.

[0036] More specifically, the process of step S5 includes:

[0037] Use the k+1,…,nth charge and discharge cycle data of the peak sequence As input, the established lithium battery health status estimation model is applied for estimation.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The ratio of the energy change within the isobaric rise to the isobaric difference is used instead of the traditional incremental analysis method, which overcomes the noise error in actual measurement and the fluctuation error caused by the differential derivative of the charging curve. The complex and violent electrochemical reactions occurring inside are characterized by external characteristics, which is conducive to the rapid extraction of features. The applicability of this method is verified under different charging rates.

[0040] Flipping layers and probability layers are used to improve GRU's reading and learning of historical and future data, reducing the mutual redundancy between model neurons to overcome this problem, improving the generalization and robustness of the model, and solving the problem of inaccurate health status estimation caused by the regeneration phenomenon that accompanies battery aging and data jumps. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a lithium battery health status estimation method based on equal pressure rise energy and improved GRU of the present invention;

[0042] Figure 2 This is an aging diagram of the health status of a lithium battery as it changes with charge and discharge cycles under different charge rate conditions provided by an embodiment of the present invention;

[0043] Figure 3 It is a constant pressure rise energy charging curve diagram under different charging rate conditions provided by an embodiment of the present invention;

[0044] Figure 4 It is an overall structural diagram of the improved GRU model provided by an embodiment of the present invention;

[0045] Figures 5a to 5c It is a comparison chart of the lithium battery health state estimation method based on equal pressure rise energy and improved GRU of the present invention and the lithium battery health state estimation results of two comparative experiments; wherein, Figure 5a This is the estimated result at a 0.5C charging rate. Figure 5b This is the estimated result at a 0.3C charging rate. Figure 5c This is the estimated result at a 0.2C charging rate;

[0046] Figures 6a to 6c It is a comparison diagram of the health state estimation error of the lithium battery health state estimation method based on equal pressure rise energy and improved GRU of the present invention and two comparative experiments; wherein, Figure 6a is the estimated error at 0.5C charging rate, Figure 6b is the estimated error at 0.3C charging rate, Figure 6c It is the estimated error at 0.2C charging rate. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be noted that the terms "upper", "lower", "inner", "outer", "front end", "rear end", "two ends", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0049] The present invention provides a lithium battery health state estimation method based on isobaric energy and improved GRU, collects health state data and charging energy data of each charge and discharge cycle of the lithium battery; extracts the isobaric interval energy sequence of each charge and discharge cycle, uses the ratio of the energy change in the isobaric energy to the isobaric difference to quickly draw the isobaric energy charging curve, and calculates the peak sequence of the isobaric energy charging curve corresponding to each charge and discharge cycle; applies the flip layer and the probability layer to improve the GRU to read and learn historical and future data, reduce the mutual redundancy between the model neurons, and construct a lithium battery health state estimation model based on the training set; applies the improved GRU health state estimation model to estimate the health state of the lithium battery and verify the data under different charging rate conditions. The present invention characterizes the complex and violent electrochemical reactions inside the battery through the external characteristics of the battery, can effectively reflect the decay and aging of the battery, and has the advantages of fast and accurate health state estimation method.

[0050] like Figure 1 As shown, the lithium battery health status estimation method based on full current and double correlation of the present invention includes the following steps.

[0051] S1. Obtain health status data and charging energy data of each charge and discharge cycle of the lithium battery;

[0052] In the embodiment of the present invention, in step S1, the health status data sequence of the lithium battery in the cyclic charge and discharge process is H1, H2, ..., H n , where the measured data Among them, H i is the battery health status of the i-th (i=1,2,K,n) charge-discharge cycle, n is the number of charge-discharge cycles, C iis the maximum discharge capacity of the lithium battery in the i-th charge and discharge process, C is the rated capacity of the lithium battery; the charging energy data are E0, E1, ..., E t , the measured energy data E j =E0+∑UIt, where E j is the charging energy at time j, E0 is the initial energy, U is the charging terminal voltage, I is the charging current, and t is the charging time.

[0053] S2. Extracting the equal-pressure rise interval energy sequence from the charging energy data of each charge-discharge cycle, drawing the equal-pressure rise energy charging curve, and calculating the peak sequence of the equal-pressure rise energy charging curve corresponding to each charge-discharge cycle;

[0054] In the embodiment of the present invention, in step S2, the energy sequence E0, E1 corresponding to the fixed voltage increase is ΔV ,…,E mΔV To quickly draw the constant pressure rise energy charging curve, calculate the peak sequence of the constant pressure rise energy charging curve as F1, F2, ..., F n ,in, F i is the maximum rate of change of the isobaric energy rise within i charging cycles, E i,aΔV , E i,(a+1)ΔV In the ith charging cycle, the charging energy of the battery in the adjacent ath and a+1th equal-voltage intervals, (a=0,…,m-1), ΔV is the interval voltage, and y is the total rising voltage during the charging process.

[0055] S3, normalizing and preprocessing the peak sequence of the equal pressure rise energy charging curve and the health state sequence as input data;

[0056] In the embodiment of the present invention, in step S3, the data is preprocessed by using the maximum and minimum normalization to normalize the data to the interval [-1, 1]; wherein the normalization formula is: x is the preprocessed data sequence, x min is the minimum value in the data sequence x, x max is the maximum value in the data sequence x.

[0057] S4. Add flipping layer and probability layer to the GRU network, improve the GRU network, and build a network estimation model for the health status of lithium batteries;

[0058] In the embodiment of the present invention, in step S4, the flip layer reverses the order of the input data sequence and reads the forward and reverse data at the same time. The flip calculation formula is as follows:

[0059] l = size(x * );

[0060]

[0061] Among them, l is a constant representing the length of the input sequence, and size is the dimension function. Hook and x * Reverse sequence of the same dimension, flip is the flip function, and dim is the dimension direction of the reverse operation.

[0062] The probability layer retains and deletes the input data according to a certain probability. The probability calculation formula is as follows:

[0063] g i = Bernoulli(p);

[0064]

[0065]

[0066] Among them, p (0<p<1) is a probability value, g i A Bernoulli distribution with probability p randomly generates a sequence of 0 and 1 with the same length as the data sequence. is the data retained according to probability, y is the output data, w is the weight of the retained data, and b is the bias.

[0067] The main data processing process in the GRU network is as follows:

[0068] z t =σ1(w z *[h t-1 , x t ]+b z )

[0069] r t =σ1(w r *[h t-1 , x t ]+b r )

[0070]

[0071]

[0072] Among them, the update gate z t Measure the degree to which the information at time t-1 is retained at time t, reset gate r t Used to control the information at time t-1 to be retained in the candidate hidden information at time t According to the updated gate state at time t and the candidate hidden state at time t, the hidden state h of the output gate is calculated. t ;w z , w r and w hare the weights of the update gate, reset gate and hidden state, respectively, and b z , b r and b h are the update gate, reset gate and hidden state bias respectively; σ1 and σ2 are sigmoid and tanh activation functions respectively, and ⊙ is the matrix element dot product.

[0073] Take the first k (k=1, ..., n-1) sample values ​​of the peak sequence and the health state sequence and As the training set, take the k+1, k+2, ..., n sample values and As a test set, a battery health state estimation model is constructed through the improved GRU network.

[0074] S5. Use the constructed lithium battery health status network estimation model to estimate the health status of the lithium battery;

[0075] In the embodiment of the present invention, in step S5, the k+1th, ..., nth charge-discharge cycle data of the peak sequence are used. As input, the established lithium battery health status estimation model is applied for estimation.

[0076] The following example is used to illustrate the process and accuracy of the lithium battery health status estimation method based on equal pressure rise energy and improved GRU proposed by the present invention.

[0077] In this example, three 18650 lithium iron phosphate (LiFePO4) batteries were subjected to 0.5C, 0.3C and 0.2C charge and discharge cycle experiments. The rated capacity was 2400mAh, and the initial discharge capacity was 2391mAh, 2376mAh, and 2355mAh, respectively. They were charged at constant currents of 1.2A, 0.72A, and 0.48A, respectively. When the battery terminal voltage rose to 4.2V, the terminal voltage was kept constant at 4.2V for constant voltage charging. Charging ended when the charging current dropped to 48mA. After 2 minutes of storage, the battery was discharged at a constant current of 2.4A until the lithium battery terminal voltage dropped to 3V, and the discharge ended. The number of charge and discharge cycles was 750. Among them, Figure 2 The health status of lithium batteries at different charge rates and the capacity decay process during charge and discharge cycles.

[0078] The specific steps are as follows:

[0079] (1) Based on the three different charging rate data collected in the laboratory, the equal pressure rise interval energy data of each charge and discharge cycle is extracted and the equal pressure rise energy charging curves at different rates are plotted. Figure 3 As shown, the peak sequence and health status data sequence of the equal pressure rise energy charging curve corresponding to 750 charge and discharge cycles are calculated.

[0080] (2) The peak data and health status data in the same charging cycle are taken as one set of data. The data of 750 sets of lithium battery charging process are divided into 1:1, that is, the data of groups 1-375 are the training set, and the data of groups 376-750 are the test set. The health status estimation model is established by improving GRU through the flip layer and probability layer. The overall network structure is as follows: Figure 4 As shown in the figure, the flip layer reads the input sequence in a longitudinal reversal, the probability of retaining and deleting the state information of neurons in the probability layer is 0.5, the number of GRU neurons is 40, and the learning rate is 0.005.

[0081] (3) The improved GRU was used to establish a lithium battery health state estimation model, and the lithium battery health state estimation was performed based on the test set. At the same time, two groups of comparative experiments were set up for three different charging rate batteries of 0.5C, 0.3C, and 0.2C, respectively, and the comparative experiments were performed with the method proposed in the present invention, as shown in Table 1. The comparison of the estimation results and the error comparison of the experiments are shown in Figures 5 and 6, respectively. Tables 2, 3, and 4 are the average absolute error and maximum error of the health state estimation results of lithium batteries with a charge rate of 0.5C, 0.3C, and 0.2C for the comparative experiments, respectively. This experiment uses the mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (R 2 ) as the evaluation indicator of the estimated model

[0082] Table 1

[0083] method enter Estimation Model The method proposed by the present invention Peak value of isobaric energy rise Improving GRU Comparison Method 1 Peak value of isobaric energy rise SVM Comparison Method 2 Peak value of isobaric energy rise ELM

[0084] Table 2

[0085] Evaluation indicators The method proposed by the present invention Comparison Method 1 Comparison Method 2 MAE(%) 0.2249 0.3364 0.3748 RMSE(%) 0.3092 0.4562 0.4934 <![CDATA[R 2 (%)]]> 99.29 97.85 97.62

[0086] Table 3

[0087] Evaluation indicators The method proposed by the present invention Comparison Method 1 Comparison Method 2 MAE(%) 0.2065 0.3138 0.3363 RMSE(%) 0.2460 0.3861 0.4088 <![CDATA[R 2 (%)]]> 99.46 97.36 97.73

[0088] Table 4

[0089] Evaluation indicators The method proposed by the present invention Comparison Method 1 Comparison Method 2 MAE(%) 0.2156 0.3604 0.4183 RMSE(%) 0.2554 0.4724 0.5267 <![CDATA[R 2 (%)]]> 98.02 94.06 93.79

[0090] From the error comparison of the health status estimation results of lithium batteries, it can be seen that the method proposed in the present invention has achieved relatively accurate SOH estimation results in the experiments of three different charge and discharge rate cycles of 0.5C, 0.3C, and 0.2C, and overcomes the battery regeneration phenomenon. At the same time, the R 2 Both are greater than 98% and higher than the R of comparison method 1 and comparison method 2 2, with a better degree of fit. The larger the degree of fit, the better the explanatory relationship between the peak characteristics of the estimated model input and the health status. This shows that the lithium battery health status estimation method based on isobaric energy rise and improved GRU proposed in the present invention can more effectively track the decay process of the battery health status, and has the characteristics of small error and high efficiency.

[0091] The present invention proposes a lithium battery health state estimation method based on isobaric energy and improved GRU. The method is: first, the health state data and charging energy data of each charge and discharge cycle of the lithium battery are obtained; then, the isobaric interval energy sequence of the charging energy data of each charge and discharge cycle is extracted, the isobaric energy charging curve is drawn, and the peak sequence of the isobaric energy charging curve corresponding to each charge and discharge cycle is calculated; finally, a flip layer and a probability layer are added to the GRU network, the GRU network is improved, and a network estimation model of the health state of the lithium battery is constructed; the constructed lithium battery health state network estimation model is used to estimate the health state of the lithium battery. The lithium battery health state estimation method based on isobaric energy and improved GRU proposed by the present invention can effectively reduce the error by considering the noise error of the incremental analysis method and the differential error after the fitting function, and adopting the ratio of the energy change in the isobaric energy to the pressure difference; the improved GRU has greatly improved the computing cost and estimation accuracy, overcomes the inaccurate estimation problem caused by the regeneration phenomenon caused by battery aging, and has great guiding significance and reference value.

[0092] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0093] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0094] The matters not described in detail in the present invention are all known technologies to those skilled in the art.

[0095] Finally, it should be noted that the above specific implementation methods are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified and replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A lithium battery health status estimation method based on equal pressure rise energy and improved GRU, characterized in that: The method comprises the following steps: S1. Obtain health status data and charging energy data of each charge and discharge cycle of the lithium battery; S2. Extracting the equal-pressure rise interval energy sequence from the charging energy data of each charge-discharge cycle, drawing the equal-pressure rise energy charging curve, and calculating the peak sequence of the equal-pressure rise energy charging curve corresponding to each charge-discharge cycle; S3, normalizing and preprocessing the peak sequence of the equal pressure rise energy charging curve and the health state sequence as input data; S4. Add flipping layer and probability layer to the GRU network, improve the GRU network, and build a network estimation model for the health status of lithium batteries; S5. Use the constructed lithium battery health status network estimation model to estimate the health status of the lithium battery. Wherein, in said step S4: The flip layer reverses the order of the input data sequence and reads both forward and reverse data at the same time. The flip calculation formula is as follows: l=size(x * ); Among them, l is a constant representing the length of the input sequence, and size is the dimension function. For x * Reverse sequence of the same dimension, flip is the flip function, dim is the dimension direction of the reverse operation; The probability layer retains and deletes the input data according to a certain probability. The probability calculation formula is as follows: g i =Bernoulli(p); Among them, p is a probability value, 0 <p<1,g i A random sequence of 0,1 with the same length as the data sequence is generated for a Bernoulli distribution with probability p, is the data retained according to probability, y is the output data, w is the weight of the retained data, and b is the bias; The main data processing process in the GRU network is as follows: z t =σ1(w z *[h t-1 ,x t ]+b z ) r t =σ1(w r *[h t-1 ,x t ]+b r ) Among them, the update gate z t Measure the degree to which the information at time t-1 is retained at time t, reset gate r t Used to control the information at time t-1 to be retained in the candidate hidden information at time t According to the updated gate state at time t and the candidate hidden state at time t, the hidden state h of the output gate is calculated. t ;w z , w r and w h are the weights of the update gate, reset gate and hidden state, respectively, and b z , b r and b h are the update gate, reset gate and hidden state bias respectively; σ1 and σ2 are sigmoid and tanh activation functions respectively, and ⊙ is the matrix element dot product; Take the first k (k=1,…,n-1) sample values ​​of the peak sequence and the health state sequence and As the training set, k=1,…,n-1, take the k+1,k+2,…,nth sample values and As a test set, a battery health state estimation model is constructed through the improved GRU network.

2. The method for estimating the health status of a lithium battery based on equal pressure rise energy and improved GRU according to claim 1, characterized in that: In step S1: The health status data sequence of lithium batteries during the charge and discharge cycle is H1, H2, ..., H n , where the measured data Among them, H i is the battery health status of the i-th charge and discharge cycle, i = 1, 2, K, n, n is the number of charge and discharge cycles, C i is the maximum discharge capacity of the lithium battery in the i-th charge and discharge process, C is the rated capacity of the lithium battery; the charging energy data are E0, E1, ..., E t , where the measured energy data E j =E0+∑UIt, where E j is the charging energy at time j, E0 is the initial energy, U is the charging terminal voltage, I is the charging current, and t is the charging time.

3. The method for estimating the health status of a lithium battery based on equal pressure rise energy and improved GRU according to claim 2, characterized in that: The process of step S2 includes: The energy sequence E0,E corresponding to the fixed voltage rise ΔV ,…,E mΔV To quickly draw the constant pressure rise energy charging curve, calculate the peak sequence of the constant pressure rise energy charging curve as F1, F2, ..., F n ,in, F i is the maximum rate of change of the isobaric energy rise within i charging cycles, E i,aΔV 、E i,(a+1)ΔV are the charging energies of the battery in the adjacent ath and a+1th equal-voltage intervals in the ith charging cycle, respectively, a=0,…,m-1, ΔV is the interval voltage, and y is the total rising voltage during the charging process.

4. The method for estimating the health status of a lithium battery based on equal pressure rise energy and improved GRU according to claim 3, characterized in that: The process of step S3 includes: The data is preprocessed by using the maximum and minimum normalization method to normalize the data to the interval [-1,1]. The normalization formula is: x is the preprocessed data sequence, x min is the minimum value in the data sequence x, x max is the maximum value in the data sequence x.

5. The method for estimating the health status of a lithium battery based on equal pressure rise energy and improved GRU according to claim 4, characterized in that: The process of step S5 includes: Use the k+1,…,nth charge and discharge cycle data of the peak sequence As input, the established lithium battery health status estimation model is applied for estimation.