A method for estimating the health status of lithium batteries based on capacity increment change curve

Abstract

The present invention discloses a method for estimating the health status of a lithium battery based on a capacity increment curve. The method comprises the following steps: acquiring lithium battery data; establishing a voltage sequence with equal voltage differences and its corresponding relationship with the charge level to construct a capacity increment curve; using a random forest method to analyze the importance of each function value on the capacity increment curve; screening function values ​​on the curve that can serve as characterization parameters; establishing a mapping relationship between the characterization parameters and the health status of the lithium battery using a neural network; and estimating the health status of the lithium battery using the neural network. The method for estimating the health status of a lithium battery of the present invention has high precision and accuracy and good versatility.

Description

Technical Field

[0001] The present invention relates to the field of lithium battery health state estimation, and in particular to a lithium battery health state estimation method based on a capacity increment change curve. Background Art

[0002] Lithium-ion batteries (LIBs) are currently widely used in electric vehicles (EVs) due to their high energy and power density, compact size, light weight, lack of memory effect, and environmental friendliness, holding considerable promise. Automotive LIBs are typically used as energy storage devices in the form of battery packs. However, varying manufacturing processes can lead to inconsistencies in LIBs, including safety, durability, and other issues. These issues can directly impact power performance or cause failures, resulting in significant inconvenience and significant repair costs, or even catastrophic consequences. For example, battery explosions due to overheating or short circuits severely limit the widespread use of LIBs in vehicles. Furthermore, LIB aging is a significant factor affecting their use and storage, potentially impacting the remaining lifespan and range of electric vehicles. Therefore, regular testing is necessary to ensure safe battery operation. Therefore, research on the estimation of the state of health (SOH) of LIBs is crucial. Accurately estimating the SOH of LIBs and obtaining this information allows for intuitive assessment of their degradation, enabling the battery management system to implement timely and effective control and management measures to improve the safety of EVs.

[0003] However, the health status estimation of lithium-ion batteries is a difficult point in the current research field. Accurate estimation of the health status is also the basis for state of charge estimation and battery balancing control. Therefore, how to provide a high-precision and high-accuracy lithium battery health status estimation method is an urgent problem that needs to be solved. Summary of the Invention

[0004] An object of the present invention is to provide a method for estimating the health status of a lithium battery based on a capacity increment change curve, and to provide at least the advantages to be described later.

[0005] The technical solutions of the present invention are as follows:

[0006] A method for estimating the health status of a lithium battery based on a capacity increment change curve comprises the following steps:

[0007] S1. Collect lithium battery data;

[0008] S2. Establishing a voltage sequence with equal voltage difference and its corresponding relationship with the charge quantity Qc-Vc to construct a capacity increment change curve;

[0009] S3. Analyze the importance of each function value of the capacity increment change curve and the health status of the lithium battery;

[0010] S4. sorting the importance of the analyzed function values ​​and screening the function values ​​on the capacity increment change curve that can be used as characterization parameters;

[0011] S5. Establishing a mapping relationship between the characterization parameters and the health status of the lithium battery through a neural network;

[0012] S6. Apply neural network to estimate the health status of lithium batteries.

[0013] Preferably, in step S2, a voltage sequence with an equal voltage difference of 0.01V is constructed according to the accuracy of the data, the minimum value Vc of the equal voltage difference voltage sequence is greater than the minimum value of the data, and the maximum value Vc is less than the maximum value of the data.

[0014] Preferably, the selection range of the equal voltage difference voltage sequence is: 3.7V≤Vc≤4.2V.

[0015] Preferably, in step S2, the corresponding relationship Qc-Vc is established as follows:

[0016] a) calculating the charge capacity Q corresponding to each voltage V of the lithium battery at each voltage point of the original data according to the ampere-hour integration method, thereby establishing a VQ relationship;

[0017] b) The VQ relationship is interpolated to obtain the charging capacity corresponding to the established equal voltage difference voltage sequence, and the corresponding relationship Qc-Vc between the equal voltage difference voltage sequence Vc and the charging capacity Qc is obtained.

[0018] Preferably, in step S2, the capacity increment change curve is constructed as follows:

[0019] Based on the corresponding relationship Qc-Vc, the IC-Vc relationship, i.e., the capacity increment curve, is calculated:

[0020]

[0021] Where IC is the capacity increment, ΔQ c is the change in charging capacity, ΔV c is the change in the voltage sequence with equal voltage difference;

[0022] The lithium battery is cycled for n times, and the IC value of the mth cycle of the capacity increment curve is taken. m and IC of the nth cycle n The difference between (1<m<n, m∈N) is used to obtain the capacity increment ΔIC for each cycle. n , and obtain ΔIC n -Vc relationship, that is, the capacity increment change curve.

[0023] Preferably, the analysis in step S3 is to analyze the importance of the function value on the capacity increment change curve and the health status of the lithium battery under each voltage sequence using a random forest method, wherein:

[0024] The random forest method is to use replacement in all training sets S A Randomly select w sub-training sets S1, S2, S3...S w , the OOB out-of-bag data of the training set that is not drawn each time is used as the test set, and w decision trees are constructed; when constructing the decision trees, F ΔICs are randomly selected from the ΔICs at all M voltages as random feature parameters to participate in the node splitting of the subsequent decision tree, where F is the largest integer less than or equal to log2(M+1);

[0025] The importance is derived in the following way: to calculate the importance of ΔIC at the x-th voltage, select the corresponding OOB out-of-bag data for the k-th decision tree (k<w) and calculate the error err of the k-th decision tree. xk1 , add random noise interference to the ΔIC at the xth voltage in the OOB out-of-bag data and then calculate the OOB out-of-bag data error err of the kth decision tree again xk2 , for the random forest method that constructs a total of w decision trees, the importance of the ΔIC at the x-th voltage to the health status of the lithium battery is expressed as: If the error err after adding the random noise xk2 A significant increase indicates that the ΔIC at this voltage has a greater impact on the results of the lithium battery health status estimation, and thus it can be seen that the importance of ΔIC at this voltage is higher.

[0026] Preferably, the function value that can be used as the characterization parameter in step S4 refers to: after sorting all the calculated function values ​​from large to small in importance, the function value of the capacity increment change curve with a larger importance value is screened out, and the characterization parameters are: 3.78V, 3.83V, 3.84V, 3.89V, 3.91V and 3.95V.

[0027] Preferably, the neural network is,

[0028] a) Establish a neuron model and select linear weighted summation to obtain the net input Net of the i-th neuron in :

[0029]

[0030] Among them, for the i-th neuron, X1, X2, ..., X i is the input of the neuron, W1, W2, ..., W iThe connection weight is used to adjust the weight ratio of each input quantity, and b is the bias;

[0031] Then Net in and θ i For comparison, the weighted input is processed by the activation function f(x) to obtain the output Y of the i-th neuron k :

[0032] Y k =f(Netin-θ i ) (4)

[0033] Among them, θ i represents the threshold of the neuron. The neuron will be activated only when the information it receives reaches the threshold.

[0034] b) a plurality of neurons constitute the input layer, hidden layer, and output layer of the neural network;

[0035] c) Neural network training: The output error is propagated back to the input layer layer by layer through the hidden layer, and the error is distributed to all units in each layer. The error signal obtained from each layer is used as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer node, the connection strength between the hidden layer node and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters corresponding to the minimum error, namely the weight and threshold, are determined.

[0036] Preferably, the mapping relationship in step S5 refers to: selecting some data on known health status as training data, using the characterization parameters obtained in step S4 as the input of the neural network, and using the health status of the lithium battery as the output for neural network training, and obtaining the mapping relationship between the characterization parameters and the health status through the neural network.

[0037] Preferably, the application of the neural network to estimate the health status of the lithium battery in step S6 is specifically: in the case of an unknown health status, by extracting a specified voltage function value on the capacity increment change curve as the input of the neural network, the neural network can output an estimated health status.

[0038] Beneficial effects

[0039] The present invention provides a lithium battery health status estimation method based on a capacity increment change curve. After analyzing the importance of function values ​​through the capacity increment change curve, function values ​​that can be used as characterization parameters are screened out, and a neural network is used to establish a mapping relationship between the characterization parameters and the lithium battery health status to estimate the lithium battery health status. The lithium battery health status estimation method of the present invention has high precision and accuracy, a simple process and high systematicity, and is suitable for actual estimation application environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flowchart of an embodiment of a method for estimating the health status of a lithium battery based on a capacity increment change curve provided by the present invention;

[0041] Figure 2 A graph showing the capacity increment variation of a battery under different cycles in one embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the neural network structure of a battery in one embodiment of the present invention;

[0043] Figure 4 This is a diagram of the neural network structure of a battery in one embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following describes the embodiments of the present invention through specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.

[0045] like Figure 1-2 As shown, the present invention provides a method for estimating the health status of a lithium battery based on a capacity increment change curve, comprising the steps of:

[0046] S1. Obtaining lithium battery data: applying the method of the present invention using 700 cycle test data of three lithium batteries;

[0047] S2. Construct a capacity increment change curve based on the data in step S1, specifically:

[0048] (1) Establishing a voltage sequence with equal voltage difference Vc: Establishing a voltage sequence with equal voltage difference of 0.01V according to data accuracy;

[0049] (2) Establish the corresponding relationship Qc-Vc between the equal voltage difference voltage sequence and the charging capacity: First, the charging capacity Q corresponding to the lithium battery charging to each voltage V is calculated at each voltage point of the original data according to the ampere-hour integration method; then establish the VQ relationship and use the cubic spline interpolation method to obtain the charging capacity corresponding to the established equal voltage difference voltage sequence, and obtain the corresponding relationship between the equal voltage difference voltage sequence Vc and the charging capacity Qc.

[0050] (3) Constructing the capacity increment curve ΔICn-Vc: Based on the obtained corresponding relationship Qc-Vc curve, the IC-Vc relationship is calculated by formula (1), that is, the capacity increment curve:

[0051]

[0052] (4) The calculated capacity increment curve is ICm Subtract ICn of the nth cycle, where 1<m<n, m∈N, to obtain the capacity increment change ΔICn of each cycle, and obtain the ΔICn-Vc relationship, that is, the capacity increment change curve.

[0053] S3. Based on the capacity increment change curve of step S2, analyze the importance of each function value on the curve: Use the random forest method to analyze the importance of the function value on the capacity increment change curve under each voltage sequence and the health status of the lithium battery. In order to calculate the importance of ΔIC at the mth voltage, for the kth decision tree, select the corresponding out-of-bag data and calculate the error err of this decision tree mk1 , randomly add noise interference to the ΔIC at the mth voltage in the out-of-bag data and then calculate the out-of-bag data error err of the kth decision tree again mk2 For a random forest with w decision trees, the importance of ΔIC under this voltage to characterize SOH can be expressed as formula (2). If random noise is added, the accuracy of out-of-bag data will drop significantly, i.e., err mk2 The sharp increase indicates that ΔIC has a greater impact on the SOH estimation results at this voltage, which further indicates that ΔIC is more important at this voltage.

[0054]

[0055] According to the selected voltage sequence, there are 50 points in total, and the importance of 50 voltage points is obtained.

[0056] S4. Based on the importance analysis in step S3, filter the function values ​​on the capacity increment change curve that can be used as characterization parameters: sort the importance of all calculated function values ​​from large to small, and filter out the function values ​​of the capacity increment change curve with an importance greater than 0.01 as characterization parameters: 3.78V, 3.83V, 3.84V, 3.89V, 3.91V and 3.95V.

[0057] S5. Establish a mapping relationship between the characterization parameters of step S4 and the health status of the lithium battery through a neural network: select data of the known health status of two batteries as training data, extract the change in the capacity increment curve at six voltages greater than 0.01V of the two batteries that are screened out as important as the input of the neural network, and train the health status of the lithium battery as the output of the neural network. Obtain the mapping relationship between the characterization parameters and the health status through the neural network, and save the trained neural network.

[0058] For the i-th neuron, X1, X2, ..., Xj are the neuron inputs, W1, W2, ..., Wj are the connection weights that adjust the weight ratio of each input, and b is the bias. Selecting linear weighted summation can obtain the net input of the i-th neuron:

[0059]

[0060] θ i Represents the threshold of the neuron. It will be activated only when the information received by the neuron reaches the threshold. in and θ i For comparison, the output obtained by processing the weighted input through the activation function f(x) is:

[0061] Y k =f(Netin-θ j ) (4)

[0062] In the neural network model, several neurons constitute the input layer, hidden layer, and output layer.

[0063] The training process propagates the output error back through the hidden layer to the input layer layer by layer, distributing the error to all units in each layer. The error signal obtained from each layer serves as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer node, the connection strength between the hidden layer node and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters (weights and thresholds) corresponding to the minimum error are determined.

[0064] S6. Applying a neural network to estimate the health status of the lithium battery: using the saved neural network to use the function values ​​of the six voltages in the capacity increment change curve of the third battery as input to estimate the health status of the third lithium battery.

Claims

1. A method for estimating the health status of a lithium battery based on a capacity increment change curve, characterized in that: The steps include: S1. Collect lithium battery data; S2. Establishing a voltage sequence with equal voltage difference and its corresponding relationship with the charge quantity Qc-Vc to construct a capacity increment change curve; S3. Analyze the importance of each function value of the capacity increment change curve and the health status of the lithium battery; S4. sorting the importance of the analyzed function values ​​and screening the function values ​​on the capacity increment change curve that can be used as characterization parameters; S5. Establishing a mapping relationship between the characterization parameters and the health status of the lithium battery through a neural network; S6. Applying neural network to estimate the health status of lithium batteries; The analysis in step S3 is to analyze the importance of the function value on the capacity increment change curve and the health status of the lithium battery under each voltage sequence using the random forest method, wherein: The random forest method is to use replacement in all training sets S A Randomly select w sub-training sets S1, S2, S3...S w , the OOB out-of-bag data of the training set that is not drawn each time is used as the test set, and w decision trees are constructed; when constructing the decision trees, F ΔICs are randomly selected from the ΔICs at all M voltages as random feature parameters to participate in the node splitting of the subsequent decision tree, where F is the largest integer less than or equal to log2(M+1); The importance is derived in the following way: to calculate the importance of ΔIC at the x-th voltage, select the corresponding OOB out-of-bag data for the k-th decision tree (k<w) and calculate the error err of the k-th decision tree. xk1 , add random noise interference to the ΔIC at the xth voltage in the OOB out-of-bag data and then calculate the OOB out-of-bag data error err of the kth decision tree again xk2 , for the random forest method that constructs a total of w decision trees, the importance of the ΔIC at the x-th voltage to the health status of the lithium battery is expressed as: If the error err after adding the random noise xk2 A significant increase indicates that the ΔIC at this voltage has a greater impact on the results of the lithium battery health status estimation, and thus it can be seen that the importance of ΔIC at this voltage is higher.

2. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 1, wherein: In the step S2, a voltage sequence with an equal voltage difference of 0.01V is constructed according to the accuracy of the lithium battery data, wherein the minimum value Vc of the equal voltage difference voltage sequence is greater than the minimum value of the data, and the maximum value Vc is less than the maximum value of the data.

3. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 2, wherein: The selection range of the equal voltage difference voltage sequence is: 3.7V≤Vc≤4.2V.

4. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 1, wherein: In step S2, the corresponding relationship Qc-Vc is established as follows: a) calculating the charge capacity Q corresponding to each voltage V of the lithium battery at each voltage point of the original data according to the ampere-hour integration method, thereby establishing a VQ relationship; b) The VQ relationship is interpolated to obtain the charging capacity corresponding to the established equal voltage difference voltage sequence, and the corresponding relationship Qc-Vc between the equal voltage difference voltage sequence Vc and the charging capacity Qc is obtained.

5. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 1, wherein: In step S2, the capacity increment change curve is constructed as follows: Based on the corresponding relationship Qc-Vc, the IC-Vc relationship, i.e., the capacity increment curve, is calculated: Where IC is the capacity increment, ΔQ c is the change in charging capacity, ΔV c is the change in the voltage sequence with equal voltage difference; The lithium battery is cycled for n times, and the IC value of the mth cycle of the capacity increment curve is taken. m and IC of the nth cycle n The difference between (1<m<n, m∈N) is used to obtain the capacity increment ΔIC for each cycle. n , and obtain ΔIC n -Vc relationship, that is, the capacity increment change curve.

6. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 1, wherein: The function value that can be used as the characterization parameter in step S4 is: after sorting all the calculated function values ​​from the most important to the least important, the function value of the capacity increment change curve with the most important value is screened out. The characterization parameter is: 3.78V, 3.83V, 3.84V, 3.89V, 3.91V, and 3.95V.

7. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 1, wherein: The neural network is, a) Establish a neuron model and select linear weighted summation to obtain the net input Net of the i-th neuron in : Among them, x i is the input of the neuron, w i is the connection weight, which is used to adjust the weight ratio of each input quantity, b is the bias; then Net in and θ i For comparison, the weighted input is processed by the activation function f(x) to obtain the output Y of the i-th neuron k : Y k =f(Net in -θ i ) (4) Among them, θ i represents the threshold of the neuron. The neuron will be activated only when the information it receives reaches the threshold. b) a plurality of neurons constitute the input layer, hidden layer, and output layer of the neural network; c) Neural network training: The output error is propagated back to the input layer layer by layer through the hidden layer, and the error is distributed to all units in each layer. The error signal obtained from each layer is used as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer node, the connection strength between the hidden layer node and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters corresponding to the minimum error, namely the weight and threshold, are determined.

8. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 7, wherein: The mapping relationship in step S5 refers to: selecting some data on known health status as training data, using the characterization parameters obtained in step S4 as the input of the neural network, and using the health status of the lithium battery as the output for neural network training, and obtaining the mapping relationship between the characterization parameters and the health status through the neural network.

9. The method for estimating the health status of a lithium battery based on a capacity increment change curve according to claim 7, wherein: The application of the neural network to estimate the health status of the lithium battery in step S6 is specifically as follows: in the case of an unknown health status, by extracting the specified voltage function value on the capacity increment change curve as the input of the neural network, the neural network can output the estimated health status.

Images