Lithium battery rul prediction method based on phase space reconstruction and multi-space attention mechanism

Abstract

The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism comprises the following steps: 1) obtaining the charging current and voltage sequence of each charge-discharge cycle of the lithium battery; 2) constructing a capacity phase space reconstruction model and inputting the obtained data to obtain a capacity phase space reconstruction matrix; 3) dividing the capacity phase space reconstruction matrix in different subspaces; 4) calculating the attention scores of different subspaces and combining them into a multi-space attention matrix; 5) constructing an offline RUL prediction model; 6) collecting the charging current sequence and charging voltage sequence of the lithium battery to be predicted, and inputting the collected data into the capacity phase space reconstruction model and the offline RUL prediction model in sequence to obtain the predicted RUL. The application uses the C-C algorithm to perform space-time embedding on the degradation capacity of the lithium battery, reconstructs the degradation mode and trend in the multi-dimensional subspace, and combines the multi-space attention mechanism to dynamically and selectively adjust the weight of different degradation rate stages, so as to improve the prediction accuracy.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of batteries, in particular to a lithium battery RUL prediction method based on phase space reconstruction and a multi-space attention mechanism. BACKGROUND

[0002] As an important supporting technology for promoting the optimization of modern energy structure, the energy storage system has become an indispensable core component of the modern energy system. As an environmentally friendly, high-energy, and recyclable energy storage solution, lithium-ion batteries have become one of the mainstream choices for current energy storage technology. However, during the degradation process of lithium batteries, the performance of the battery gradually decreases, and the degradation rate at different stages usually shows significant differences. For example, the degradation rate is relatively gentle in the initial stage, while in the middle and later stages, it may show accelerated or even nonlinear dramatic changes. This multi-stage, multi-rate degradation behavior makes it difficult for one-dimensional capacity data-based analysis to fully capture the complex dynamic characteristics and degradation laws at each stage, thereby adversely affecting the accuracy and reliability of the remaining useful life (RUL) prediction.

[0003] Therefore, there is an urgent need for a nonlinear dynamics analysis technique to reveal the dynamic characteristics of battery capacity degradation and a dynamic weight distribution mechanism for different degradation stages to improve the adaptability of the RUL prediction model in complex nonlinear degradation environments. This is of great significance for the life optimization management of energy storage systems. SUMMARY

[0004] The purpose of the application is to provide a lithium battery RUL prediction method based on phase space reconstruction and a multi-space attention mechanism, comprising the following steps:

[0005] 1) Obtain the charging current sequence and charging voltage sequence of each charge-discharge cycle of the lithium battery.

[0006] 2) Construct a capacity phase space reconstruction model, and input the obtained charging current sequence and charging voltage sequence into the capacity phase space reconstruction model to obtain a capacity phase space reconstruction matrix.

[0007] 3) According to the capacity degradation rate characteristics of the capacity phase space reconstruction matrix, divide the capacity phase space reconstruction matrix into different subspaces.

[0008] 4) Apply a multi-head attention mechanism to calculate the attention scores of different subspaces, and combine the attention scores of different subspaces into a multi-space attention matrix through a fully connected layer.

[0009] 5) Pass the multi-space attention matrix to a recurrent neural network GRU, and construct an offline RUL prediction model with the capacity phase space reconstruction matrix as input and the remaining useful life RUL of the lithium battery as output.

[0010] 6) Collect the charging current sequence and the charging voltage sequence of the lithium battery to be predicted, and input them into the capacity phase space reconstruction model to obtain the capacity phase space reconstruction matrix of the lithium battery to be predicted.

[0011] 7) Input the capacity phase space reconstruction matrix of the lithium battery to be predicted into the offline RUL prediction model to obtain the remaining useful life RUL of the lithium battery to be predicted.

[0012] Further, in step 2), the steps of obtaining the capacity phase space reconstruction matrix are as follows:

[0013] 2.1) Smooth the data input into the capacity phase space reconstruction model to obtain processed data.

[0014] 2.2) Calculate the battery capacity based on the processed data.

[0015] 2.3) Determine the optimal embedding dimension and time delay based on the battery capacity using the C-C algorithm, and construct the capacity phase space reconstruction matrix.

[0016] Further, the smoothing processing adopts sliding mean filtering, as shown below:

[0017]

[0018] In the formula, i is the charge and discharge cycle index. t represents the time. p is the data collection number. L is the size of the sliding window. I i,t , V i,t I

[0019] Further, the battery capacity is as follows:

[0020]

[0021] In the formula, i is the charge and discharge cycle index, n is the total cycle number. t represents the time. p is the data collection number, and e is the end time. I nom V i Q

[0022] Further, the battery capacity is also standardized by normalization processing, and the calculation formula of the normalization processing is as follows:

[0023]

[0024] wherein X * is the normalized data, X is the data to be normalized, X max , X min are the maximum and minimum values in the sequence X, respectively.

[0025] Further, in step 2.3), the steps of determining the optimal embedding dimension and time delay using the C-C algorithm are as follows:

[0026] 2.3.1) Based on the battery capacity, the cross-correlation integral is calculated as follows:

[0027]

[0028] wherein C(m, n, r, τ) is the cross-correlation integral. i, j are both the charge and discharge cycle indexes, and n is the total cycle number. Q i , Q j represent the charge capacity of the i-th and j-th cycles, respectively. m represents the embedding dimension, M is the number of embedding points in the m-dimensional space, and M = n - (m - 1)τ, τ represents the time delay, and r represents the spatial distance.

[0029] wherein the piecewise function F is as follows:

[0030]

[0031] 2.3.2) According to the cross-correlation integral and the time delay average strategy, the nonlinear feature function S(m, n, r, τ) is calculated as follows:

[0032]

[0033] wherein S(m, n, r, τ) is the nonlinear feature function, S represents the time delay index, and both represent the cross-correlation integral.

[0034] 2.3.3) The nonlinear feature function S(m, n, r, τ) is converted into the minimum position of the rate of change of the arbitrary spatial distance r, and the optimal time delay τ is derived.

[0035] The minimum position of the rate of change of the arbitrary spatial distance r is as follows:

[0036] ΔS(m, τ) = max{S(m, r j , τ)} - min{S(m, r j , τ)} (7)

[0037] wherein ΔS(m, τ) represents the minimum position of the rate of change of the arbitrary spatial distance r, S(m, r j , τ) represents the nonlinear feature function of the arbitrary spatial distance r when n is infinite.

[0038] 2.3.4) The optimal embedding dimension is obtained based on the time delay window, as shown below:

[0039] τ window =(m-1)τ (8)

[0040]

[0041] In the formula, τ window This is the time delay window. It is the average value of the locations where the rate of change of any spatial distance r is the minimum. Let r be the average value of the nonlinear characteristic function of any spatial distance r.

[0042] Furthermore, the capacity phase space reconstruction matrix is ​​shown below:

[0043]

[0044] In the formula, i is the charge / discharge cycle index, n is the total number of cycles, and m represents the embedding dimension. Q i Q represents the charging capacity in the i-th cycle. R Q No These represent incremental capacity with and without regeneration capacity, respectively.

[0045] Furthermore, the regenerative incremental capacity Q R and non-regenerative incremental capacity Q No As shown below:

[0046]

[0047] In the formula, q represents the upper limit of the regenerative incremental capacity.

[0048] Furthermore, the multi-space attention matrix is ​​as follows:

[0049]

[0050] HAM=σ(MH*W space +b space (14)

[0051] In the formula, Attention is the single-head attention score, A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, and d is the query vector matrix. k It is the scaling factor, Concat indicates matrix concatenation, i r For single-head attention indexing, Indicates single-head attention i r The hidden state, h is the number of single-head attention. Query vector matrix for single-head attention i r Key vector matrix for single-head attention i Query vector matrix for single-head attention i r Key vector matrix for single-head attention i Key vector matrix for single-head attention i r Key vector matrix for single-head attention i, σ is an activation function, W, W space are subspace weight matrix, multi-space weight matrix, respectively, b space is multi-space bias. HAM is multi-space attention matrix. MH is the spliced single-head attention matrix.

[0052] Further, the offline RUL prediction model is as follows:

[0053] z t = σ1(w z *[h t-1 , HAM t ]+b z ) (15)

[0054] r t = σ1(w r *[h t-1 , HAM t ]+b r ) (16)

[0055]

[0056] In the formula, HAM t is the multi-space attention matrix input at time t. z t , r t , h t respectively represent the update gate, reset gate, candidate hidden state, hidden state output at time t. h t-1 is the output of the hidden state at time t-1. σ1, σ2 are sigmoid function, hyperbolic tangent activation function, respectively, and is matrix element point product. w z , w r , w h are the weight matrices of the update gate, reset gate, and hidden state, respectively. b z , b r , b h are the biases of the update gate, reset gate, and hidden state, respectively.

[0057] The technical effect of the present application is self-evident, and the present application provides a lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism. The method uses the C-C algorithm to perform space-time embedding on the degradation capacity of the lithium battery, reconstructs the degradation mode and trend in the multidimensional subspace, and combines the multi-space attention mechanism to dynamically and selectively adjust the weight of different degradation rate stages to improve the prediction accuracy.

[0058] The present application adopts the C-C algorithm to perform phase space reconstruction on one-dimensional sequence data, which can more comprehensively represent the dynamic behavior mode and degradation rate difference of the lithium battery in different degradation stages, thereby providing a rich data basis for subsequent feature extraction and analysis.

[0059] In the reconstructed high-dimensional space, the present application further applies the multi-space attention mechanism to identify the key stage that has a significant influence on the degradation characteristics, and preferentially gives it a higher weight, effectively balancing the feature expression strength of different degradation stages, and enhancing the robustness and reliability of RUL prediction in short-term and long-term applications.

[0060] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism provided by the present application can effectively capture the dynamic characteristics of the battery in different degradation stages, exhibit strong robustness and high prediction accuracy, and provide strong support for lithium battery health management and life prediction in actual applications. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a flowchart of the lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism provided by the present application;

[0062] Figure 2 is the B0005 and B0006 battery capacity aging curve of the NASA battery data set provided by the present application;

[0063] Figure 3 is the smoothed curve of the B0006 battery capacity calculation provided by the present application;

[0064] Figure 4 is the RUL prediction comparison experiment result graph using the first 30 cycle capacity data of the B0006 battery provided by the present application;

[0065] Figure 5 is the RUL prediction comparison experiment result graph using the first 50 cycle capacity data of the B0006 battery provided by the present application;

[0066] Figure 6 is the RUL prediction comparison experiment result graph using the first 70 cycle capacity data of the B0006 battery provided by the present application. DETAILED DESCRIPTION

[0067] The application will be further described below in connection with the embodiments, but should not be understood as limiting the above-mentioned subject matter of the application only to the following embodiments. Various replacements and changes can be made according to the ordinary technical knowledge and common means in the art without departing from the above-mentioned technical idea of the application, and all of them should be included in the protection scope of the application.

[0068] Embodiment 1

[0069] Reference Figures 1 to 6 The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism includes the following steps:

[0070] 1) Obtain the charging current sequence and the charging voltage sequence in each charge-discharge cycle of the lithium battery.

[0071] 2) Construct a capacity phase space reconstruction model, and input the obtained charging current sequence and charging voltage sequence into the capacity phase space reconstruction model to obtain a capacity phase space reconstruction matrix.

[0072] 3) According to the capacity degradation rate characteristics of the capacity phase space reconstruction matrix, the capacity phase space reconstruction matrix is divided into different subspaces.

[0073] 4) Apply a multi-head attention mechanism to calculate the attention scores of different subspaces, and combine the attention scores of different subspaces into a multi-space attention matrix through a fully connected layer.

[0074] 5) Pass the multi-space attention matrix to a recurrent neural network GRU, and construct an offline RUL prediction model with the capacity phase space reconstruction matrix as the input and the remaining useful life RUL of the lithium battery as the output.

[0075] 6) Collect the charging current sequence and the charging voltage sequence of the lithium battery to be predicted, and input them into the capacity phase space reconstruction model to obtain the capacity phase space reconstruction matrix of the lithium battery to be predicted.

[0076] 7) Input the capacity phase space reconstruction matrix of the lithium battery to be predicted into the offline RUL prediction model to predict the remaining useful life RUL of the lithium battery to be predicted.

[0077] Embodiment 2

[0078] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content is seen in embodiment 1, further, in step 2), the steps of obtaining the capacity phase space reconstruction matrix are as follows:

[0079] 2.1) Smooth the data input into the capacity phase space reconstruction model to obtain processed data.

[0080] 2.2) Based on the processed data, calculate the battery capacity.

[0081] 2.3) Based on the battery capacity, determine the optimal embedding dimension and time delay using the C-C algorithm, and construct the capacity phase space reconstruction matrix.

[0082] Embodiment 3:

[0083] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 2, further, the smoothing processing adopts sliding mean filtering, as shown below:

[0084]

[0085] In the formula, i is the charge and discharge cycle index. t represents the time. p is the data acquisition number. L is the size of the sliding window. respectively represent the current and voltage data at time t in the i th cycle after smoothing processing. i,t , V i,t respectively represent the current and voltage data at time t in the i th cycle after smoothing processing.

[0086] Embodiment 4:

[0087] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 3, further, the battery capacity is as follows:

[0088]

[0089] In the formula, i is the charge and discharge cycle index, n is the total cycle number. t represents the time. p is the data acquisition number, and e is the end time. respectively represent the current and voltage data at time t in the i th cycle after smoothing processing. nom is the nominal voltage. Q i represents the charge capacity of the i th cycle.

[0090] Embodiment 5:

[0091] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 4, further, the battery capacity also adopts normalization processing to standardize the data, and the calculation formula of the normalization processing is as follows:

[0092]

[0093] In the formula, X *For the normalized data, X is the data to be normalized, X max , X min are the maximum and minimum values in the sequence X, respectively.

[0094] Example 6:

[0095] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content is any one of embodiments 1 to 5, further, in step 2.3), the steps of determining the optimal embedding dimension and time delay using C-C algorithm are as follows:

[0096] 2.3.1) Based on the battery capacity, the cross-correlation integral is calculated as follows:

[0097]

[0098] In the formula, C(m, n, r, τ) is the cross-correlation integral. i, j are both charge and discharge cycle indexes, and n is the total cycle number. Q i , Q j represent the charge capacity of the i-th and j-th cycles, respectively. m represents the embedding dimension, M is the number of embedding points in the m-dimensional space, and M = n-(m-1)τ, τ represents the time delay, and r represents the spatial distance.

[0099] Where the piecewise function F is as follows:

[0100]

[0101] 2.3.2) According to the cross-correlation integral and the time delay average strategy, the nonlinear feature function is calculated as follows:

[0102]

[0103] In the formula, S(m, n, r, τ) is the nonlinear feature function, S represents the time delay index, and both represent the cross-correlation integral.

[0104] 2.3.3) Convert the nonlinear feature function S(m, n, r, τ) to the minimum position of the change rate of the arbitrary spatial distance r, and derive the optimal time delay τ.

[0105] The minimum position of the change rate of the arbitrary spatial distance r is as follows:

[0106] ΔS(m, τ) = max{S(m, r j , τ)} - min{S(m, r j , τ)} (7)

[0107] where ΔS(m, τ) represents the minimum position of the rate of change of any spatial distance r, S(m, r j , τ) represents the nonlinear characteristic function of any spatial distance r when n is infinite.

[0108] 2.3.4) The optimal embedding dimension is solved based on the delay window, as follows:

[0109] τ window = (m-1)τ (8)

[0110]

[0111] where τ window is the delay window. is the average value of the minimum position of the rate of change of any spatial distance r. is the average value of the nonlinear characteristic function of any spatial distance r.

[0112] Example 7:

[0113] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 6, further, the capacity phase space reconstruction matrix is as follows:

[0114]

[0115] where i is the charge and discharge cycle index, n is the total cycle number, m represents the embedding dimension, Q i represents the charge capacity of the i-th cycle. Q R , Q No respectively represent the regenerated incremental capacity, no regenerated incremental capacity.

[0116] Example 8:

[0117] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 7, further, the regenerated incremental capacity Q R and no regenerated incremental capacity Q No are as follows:

[0118]

[0119] where q is the upper limit of the regenerated incremental capacity.

[0120] Example 9:

[0121] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 1 to 8, further, the multi-space attention matrix is as follows:

[0122]

[0123] HAM = σ (MH * W space +b space ) (14)

[0124] wherein Attention is the single-head attention score, A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, d k is the scaling factor, Concat denotes the matrix concatenation, i r is the single-head attention index, denotes the hidden state of the single-head attention i r , h is the number of single-head attentions, is the query vector matrix of the single-head attention i r , is the key vector matrix of the single-head attention i r , is the key vector matrix of the single-head attention i r , σ is the activation function, W, W space are the subspace weight matrix, the multi-space weight matrix, respectively, b space is the multi-space bias. HAM is the multi-space attention matrix. MH is the single-head attention matrix after splicing.

[0125] Embodiment 10:

[0126] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content is any one of embodiments 1 to 9, further, the offline RUL prediction model is as follows:

[0127] z t = σ1(w z *[h t-1 ,HAM t ]+b z ) (15)

[0128] r t = σ1(w r *[h t-1 ,HAM t ]+b r ) (16)

[0129]

[0130] wherein HAM t is the multi-space attention matrix input at time t. z t , r t , ht respectively represent the update gate, reset gate, candidate hidden state, and output of the hidden state at time t. t-1 respectively represent the output of the hidden state at time t-1. σ1, σ2 are sigmoid function, hyperbolic tangent activation function respectively, and is matrix element dot product. z r h are weight matrices of the update gate, reset gate, and hidden state respectively. z r h are biases of the update gate, reset gate, and hidden state respectively.

[0131] Embodiment 11:

[0132] Referring to Figures 1 to 6 , the lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism includes the following steps:

[0133] 1) Obtain the charging current sequence and charging voltage sequence of each charge-discharge cycle of the lithium battery.

[0134] 2) Construct a capacity phase space reconstruction model, and input the obtained charging current sequence and charging voltage sequence into the capacity phase space reconstruction model to obtain a capacity phase space reconstruction matrix.

[0135] 3) According to the capacity degradation rate characteristics of the capacity phase space reconstruction matrix, the capacity phase space reconstruction matrix is divided into different subspaces.

[0136] 4) Apply a multi-head attention mechanism to calculate the attention scores of different subspaces, and combine the attention scores of different subspaces into a multi-space attention matrix through a fully connected layer.

[0137] The multi-space attention mechanism adjusts the dynamic relationship between different subspaces and focuses on key periods and degradation patterns that have a significant impact on capacity degradation, thereby effectively improving the accuracy and robustness of the model in short-term and long-term prediction tasks.

[0138] 5) Pass the multi-space attention matrix to a recurrent neural network GRU to construct an offline RUL prediction model with the capacity phase space reconstruction matrix as input and the remaining useful life RUL of the lithium battery as output.

[0139] 6) Collect the charging current sequence and charging voltage sequence of the lithium battery to be predicted, and input them into the capacity phase space reconstruction model to obtain the capacity phase space reconstruction matrix of the lithium battery to be predicted.

[0140] ​​​​7) input the capacity phase space reconstruction matrix of the lithium battery to be predicted into the offline RUL prediction model, and predict the remaining useful life RUL of the lithium battery to be predicted.

[0141] Embodiment 12:

[0142] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of which is seen from any one of embodiments 11 to 13, further, the capacity is as follows:

[0143] 2.1) smoothing the data input into the capacity phase space reconstruction model to obtain processed data.

[0144] 2.2) calculating the battery capacity based on the processed data.

[0145] 2.3) determining the optimal embedding dimension and time delay based on the battery capacity, and constructing the capacity phase space reconstruction matrix by using the C-C algorithm.

[0146] Embodiment 13:

[0147] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of which is seen from any one of embodiments 11 to 13, further, the smoothing processing adopts sliding mean filtering, as shown below:

[0148]

[0149] In the formula, i is the charge and discharge cycle index. t represents the time. p is the data collection number. L is the size of the sliding window. I i,t , V i,t I nom represent the current and voltage data at time t in the i th cycle after smoothing processing respectively.

[0150] Embodiment 14:

[0151] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of which is seen from any one of embodiments 11 to 13, further, the battery capacity is as follows:

[0152]

[0153] In the formula, i is the charge and discharge cycle index, n is the total cycle number. t represents the time. p is the data collection number, and e is the end time. I nom is the nominal voltage. Q iQ i represents the charge capacity of the i-th cycle.

[0154] Embodiment 15:

[0155] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 11 to 14, further, the battery capacity also adopts a normalization processing to standardize the data, and maps the data to the interval [-1, 1] to ensure the uniform distribution of the data in the interval, and the calculation formula of the normalization processing is as follows:

[0156]

[0157] In the formula, X * is the normalized data, X is the data to be normalized, X max , X min are the maximum value and the minimum value in the sequence X respectively.

[0158] Embodiment 16:

[0159] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 11 to 15, further, in step 2.3), the steps of determining the optimal embedding dimension and time delay by using C-C algorithm are as follows:

[0160] 2.3.1) Based on the battery capacity, the cross-correlation integral is calculated as follows:

[0161]

[0162] In the formula, C(m, n, r, τ) is the cross-correlation integral. i, j are both the charge and discharge cycle indexes, and n is the total cycle number. Q i , Q j represent the charge capacity of the i-th and j-th cycles respectively. m represents the embedding dimension, M is the number of embedding points in the m-dimensional space, and M = n-(m-1)τ, τ represents the time delay, and r represents the spatial distance.

[0163] Wherein, the piecewise function F is as follows:

[0164]

[0165] 2.3.2) According to the cross-correlation integral and the time delay average strategy, the nonlinear feature function S(m, n, r, τ) is calculated as follows:

[0166]

[0167] In the formula, S(m, n, r, τ) is the nonlinear feature function, S represents the time delay index, and all represent cross-correlation integrals.

[0168] 2.3.3) Convert the nonlinear characteristic function S(m, n, r, τ) to the minimum position of the rate of change of the arbitrary spatial distance r, and derive the optimal time delay τ.

[0169] Assuming that m and τ are fixed, when n approaches infinity, S(m, n, r, τ) can be derived from the cross-correlation integral C(m, n, r, τ) at all spatial distances r. Therefore, the optimal delay τ can be the minimum position of the rate of change of the arbitrary relative spatial distance r converted by S(m, n, r, τ).

[0170] The minimum position of the rate of change of the arbitrary spatial distance r is as follows:

[0171] ΔS(m, τ) = max{S(m, r j , τ)} - min{S(m, r j , τ)} (7)

[0172] In the formula, ΔS(m, τ) represents the minimum position of the rate of change of the arbitrary spatial distance r, S(m, r j , τ) represents the nonlinear characteristic function of the arbitrary spatial distance r when n is infinite.

[0173] 2.3.4) According to the time delay window At the minimum value, equal to the time delay τ, the embedding dimension m can be solved by τ window = (m-1) τ. The optimal embedding dimension is solved based on the time delay window, as follows:

[0174] τ window = (m-1) τ (8)

[0175]

[0176] In the formula, τ wi nd ow is the time delay window. is the average value of the minimum position of the rate of change of the arbitrary spatial distance r. is the average value of the nonlinear characteristic function of the arbitrary spatial distance r.

[0177] Embodiment 17:

[0178] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content is any one of embodiments 11 to 16, further, the capacity phase space reconstruction matrix is as follows:

[0179]

[0180] In the formula, i is the charge / discharge cycle index, n is the total number of cycles, m represents the embedding dimension, and Q... i Q represents the charging capacity in the i-th cycle. R Q No These represent incremental capacity with and without regeneration capacity, respectively.

[0181] Example 18:

[0182] A lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical contents of which are described in any one of Examples 11 to 17, further, the regenerated incremental capacity Q R and non-regenerative incremental capacity Q No As shown below:

[0183]

[0184] In the formula, q represents the upper limit of the regenerative incremental capacity.

[0185] Example 19:

[0186] The lithium battery RUL prediction method based on phase space reconstruction and multi-spatial attention mechanism, the main technical contents of which are described in any one of Examples 11 to 18, and further, the multi-spatial attention matrix is ​​as follows:

[0187]

[0188] HAM=σ(MH*W space +b space (14)

[0189] In the formula, Attention is the single-head attention score, A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, and d is the query vector matrix. k It is the scaling factor, Concat indicates matrix concatenation, i r For single-head attention indexing, Indicates single-head attention i r The hidden state, h is the number of single-head attention. For single-head attention i r The query vector matrix, For single-head attention i r The key vector matrix, For single-head attention i r The key vector matrix, σ is the activation function, W, W space These are the subspace weight matrix and the multispace weight matrix, respectively. spaceis a multi-space bias. HAM is a multi-space attention matrix. MH is a single-head attention matrix after splicing.

[0190] Embodiment 20:

[0191] The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, the main technical content of any one of embodiments 11 to 19, further, the offline RUL prediction model is as follows:

[0192] z t = σ1(w z *[h t-1 ,HAM t ]+b z ) (15)

[0193] r t = σ1(w r *[h t-1 ,HAM t ]+b r ) (16)

[0194]

[0195] In the formula, HAM t is a multi-space attention matrix input at time t. z t , r t , h t respectively represent the update gate, reset gate, candidate hidden state, and hidden state output at time t. h t-1 is the output of the hidden state at time t-1. σ1 and σ2 are sigmoid function and hyperbolic tangent activation function respectively, and is matrix element point product. w z , w r , w h are the weight matrices of the update gate, reset gate, and hidden state respectively. b z , b r , b h are the biases of the update gate, reset gate, and hidden state respectively.

[0196] Embodiment 21:

[0197] Referring to Figures 1 to 6 , the lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism includes the following steps:

[0198] S1, collect the charge current and voltage sequence of each charge and discharge cycle of the lithium battery, and smooth the data by using sliding mean filtering.

[0199] S2, calculate the battery capacity and normalize the data according to the processed current and voltage sequences.

[0200] S3, determine the optimal embedding dimension and time delay mapping to high-dimensional phase space using C-C algorithm, and obtain the capacity phase space reconstruction matrix.

[0201] S4, according to the capacity degradation rate characteristics of the capacity phase space reconstruction matrix, it is divided into different subspaces.

[0202] S5, independently apply multi-head attention mechanism to calculate attention score in different subspaces, and combine into multi-space attention matrix through full connection layer to pass to GRU to establish offline RUL prediction model.

[0203] S6, execute steps S1-S3 on the current and voltage sequence data of unknown RUL lithium battery, and input to step S6 for online lithium battery RUL prediction.

[0204] The specific content of step S1 is as follows:

[0205] The charging current and voltage sequence in each charge-discharge cycle are [I i,1 ,I i,2 ,…,I i,t …,I i,e ] and [V i,1 ,V i,2 ,…,V i,t …,V i,e ], wherein I i,t is the collection value of current at time t in the ith cycle, and e is the end time. V i,t is the collection value of voltage at time t in the ith cycle, and e is the end time. Then the sliding mean filter is used to smooth the current and voltage sequence data respectively to obtain wherein, and are the smoothed values of current and voltage at time t in the ith cycle, L is the size of sliding window, and p is the data collection number.

[0206] The specific content of step S2 is as follows:

[0207] The battery capacity can be calculated according to the smoothed current and voltage sequences as wherein, Q i is the charging capacity of i cycles, n is the total cycle number, and V nom is the nominal voltage. Further, the data is standardized by maximum and minimum value normalization, which maps the data to the interval [-1, 1] to ensure uniform distribution of data in the interval. The normalization formula is X is the preprocessed data sequence, Xmax and Xmin are the maximum and minimum values of the sequence X respectively.

[0208] The specific content of step S3 is as follows:

[0209] Suppose a one-dimensional capacity sequence Q n = [Q1, Q2, … Q i , …, Q n ], n is the total number of cycles. The optimal embedding dimension and time delay calculated by the C-C algorithm.

[0210] First, calculate the cross-correlation integral where M = n - (m - 1)τ is the number of embedding points in m-dimensional space, r represents the spatial distance, τ represents the reconstruction time delay, and the piecewise function

[0211] Then, according to the cross-correlation integral C(m, n, r, τ) and the time delay average strategy, the nonlinear characteristic function S(m, n, r, τ) is obtained Suppose m and τ are fixed, when n approaches infinity, S(m, n, r, τ) can be derived from the cross-correlation integral C(m, n, r, τ) at all spatial distances r, and is equal to 0. Therefore, the optimal delay τ can be the minimum position of the change rate of S(m, n, r, τ) converted to any relative spatial distance r, that is, ΔS(m, τ) = max{S(m, r j , τ)} - min{S(m, r j , τ)}.

[0212] Finally, according to the time delay window At the minimum value, equal to the time delay τ, the embedding dimension m can be solved by τ window = (m - 1)τ.

[0213] The specific content of step S4 is as follows:

[0214] According to the optimal embedding dimension and time delay calculated in step 3, the one-dimensional capacity sequence Q n = [Q1, Q2, … Q i , …, Q n ] of length n is reconstructed into an m×τ-dimensional capacity matrix, and is divided according to the incremental capacity value: capacity degradation with regeneration and without regeneration. The reconstructed capacity matrix is as follows:

[0215]

[0216] wherein The capacity regeneration phenomenon is that the battery capacity will experience a temporary recovery, an increase compared to the previous cycle, usually between 0% and 5%.

[0217] The specific content of step S5 is as follows:

[0218] According to the subspaces with different degradation rates divided in step S4, a multi-head attention mechanism is independently applied to calculate the attention score matrix of each subspace, and combined into a multi-space attention matrix HAM through a full connection layer.

[0219] The multi-space attention mechanism adjusts the dynamic relationship between different subspaces and focuses on the key period and degradation mode which has a significant impact on capacity degradation, thereby effectively improving the accuracy and robustness of the model in short-term and long-term prediction tasks. The calculation process is as follows:

[0220]

[0221] MH=Concat(head1,...,head i ,…,head h )W

[0222] HAM=σ(MH*W space +b space )

[0223] Where A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, dk is the scaling factor, Concat represents matrix connection, head i =Attention(Q i ,K i ,V i ),i∈{1,2,…,h} is the hidden state of single-head attention, h is the number of single-head attention, σ is the activation function, W and Wspace are subspace weight matrix and multi-space weight matrix respectively, and bspace is the multi-space bias.

[0224] Its multi-space attention matrix HAM is passed to the GRU to establish the offline RUL prediction model process as follows:

[0225] z t =σ1(w z *[h t-1 ,HAM t ]+b z )

[0226] r t =σ1(w r *[h t-1 ,HAM t ]+b r )

[0227]

[0228] Where HAM tInput multi-space attention matrix at time t. z t t t The update gate, reset gate, candidate hidden state and hidden state output at time t, respectively. σ1 and σ2 are sigmoid and hyperbolic tangent activation functions, respectively, and ⊙ is the matrix element dot product. wz, wr and wh are the weight matrices of the update gate, reset gate and hidden layer, respectively. bz, br and bh are the biases of the update gate, reset gate and hidden state, respectively.

[0229] The specific content of step S6 is as follows:

[0230] Obtain the phase space reconstruction matrix using the current and voltage sequence data of the lithium battery with unknown RUL As an input of the offline RUL prediction model to predict the RUL of the unknown lithium battery.

[0231] Embodiment 22:

[0232] Referring to Figures 1 to 6 , the lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism includes the following steps:

[0233] S1, collect the charging current and voltage sequence in each charge-discharge cycle of the lithium battery, and smooth the data by using the sliding mean filtering;

[0234] In this embodiment, in step S1, the charging current and voltage sequence in each charge-discharge cycle are [I i,1 ,I i,2 ,…,I i,t …,I i,e ] and [V i,1 ,V i,2 ,…,V i,t …,V i,e ], respectively, where I i,t is the collection value of the current at time t in the ith cycle, and e is the end time; V i,t is the collection value of the voltage at time t in the ith cycle, and e is the end time. Then, the current and voltage sequence data are smoothed by using the sliding mean filtering to obtain wherein, and are the smoothed values of the current and voltage at time t in the ith cycle, respectively, L is the size of the sliding window, and p is the data collection number.

[0235] S2, calculate the battery capacity according to the processed current and voltage sequence and normalize the data;

[0236] ​​​In the embodiment, in step S2, the battery capacity can be calculated according to the smoothed current and voltage sequences where Q i is the charge capacity of i cycles, n is the total cycle number, V nom is the nominal voltage. Further, the data is standardized by using the maximum-minimum value normalization to map the data to the interval [-1, 1] to ensure uniform distribution of the data in the interval; wherein the normalization formula is X is the preprocessed data sequence, Xmax and Xmin are the maximum and minimum values in the sequence X respectively.

[0237] S3, determine the optimal embedding dimension and time delay mapping to high-dimensional phase space by using C-C algorithm to obtain the capacity phase space reconstruction matrix;

[0238] In the embodiment, in step S3, assuming that the one-dimensional capacity sequence Q n = [Q1, Q2, … Q i , …, Q n ], n is the total cycle number. The optimal embedding dimension and time delay are calculated by using C-C algorithm.

[0239] First, calculate the cross-correlation integral where M = n-(m-1)τ is the number of embedding points in the m-dimensional space, r represents the spatial distance, τ represents the reconstruction time delay, and the piecewise function

[0240] Then, the nonlinear feature function S(m, n, r, τ) is obtained according to the cross-correlation integral C(m, n, r, τ) and the time delay average strategy Assuming that m and τ are fixed and unchanged, when n approaches infinity, S(m, n, r, τ) can be derived from the cross-correlation integral C(m, n, r, τ) to be equal to 0 at all spatial distances r. Therefore, the optimal delay τ can be the minimum position of the change rate of S(m, n, r, τ) converted to any relative spatial distance r, i.e. ΔS(m, τ) = max{S(m, r j , τ)}-min{S(m, r j , τ)}.

[0241] Finally, according to the time delay window At the minimum value, τ equals the time delay τ, and the embedding dimension m can be solved by τ window =(m-1)τ.

[0242] S4, according to the capacity degradation rate characteristics of the capacity phase space reconstruction matrix, it is divided into different subspaces;

[0243] In this embodiment, in step S4, the one-dimensional capacity sequence Q of length n calculated in step 3 is reconstructed into a capacity matrix of m x τ dimensions according to the optimal embedding dimension and time delay, and is divided into capacity degradation with regeneration and capacity degradation without regeneration according to the incremental capacity value. The reconstructed capacity matrix is as follows: n = [Q1, Q2, … Q i ,…,Q n ]

[0244]

[0245] wherein, The capacity regeneration phenomenon is that the battery capacity will experience a temporary recovery, an increase compared to the previous cycle, usually between 0% and 5%.

[0246] S5, independently applying a multi-head attention mechanism to calculate attention scores in different subspaces, and combining them into a multi-space attention matrix through a fully connected layer, and passing it to the GRU to establish an offline RUL prediction model;

[0247] In this embodiment, in step S5, according to the subspaces with different degradation rates divided in step S4, a multi-head attention mechanism is independently applied to calculate the attention score matrix of each subspace, and a multi-space attention matrix HAM is combined through a fully connected layer.

[0248] The multi-space attention mechanism adjusts the dynamic relationship between different subspaces and focuses on key periods and degradation patterns that have a major impact on capacity degradation, thereby effectively improving the accuracy and robustness of the model in short-term and long-term prediction tasks. The calculation process is as follows:

[0249]

[0250] MH = Concat (head1, …, head i ,…,head h )W

[0251] HAM = σ (MH * W space +b space )

[0252] wherein, A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, dk is the scaling factor, Concat represents matrix concatenation, head i = Attention (Q i ,K i ,V i), i e {1, 2,..., h} is the hidden state of the single-head attention, h is the number of single-head attention, σ is the activation function, W and Wspace are the subspace weight matrix and the multi-space weight matrix respectively, bspace is the multi-space bias.

[0253] The multi-space attention matrix HAM thereof is delivered to the GRU to establish the offline RUL prediction model process as follows:

[0254] z t = σ1(w z *[h t-1 ,HAM t ]+b z )

[0255] r t = σ1(w r *[h t-1 ,HAM t ]+b r )

[0256]

[0257] wherein HAM t is the multi-space attention matrix input at t time; z t , r t , and h t represent the update gate, the reset gate, the candidate hidden state and the hidden state output at t time respectively; σ1 and σ2 are sigmoid and hyperbolic tangent activation functions respectively, and is the matrix element dot product; wz, wr and wh are the weight matrices of the update gate, the reset gate and the hidden layer respectively; bz, br and bh are the biases of the update gate, the reset gate and the hidden state respectively.

[0258] S6, the current and voltage sequence data of the unknown RUL lithium battery are executed in steps S1-S3 and input to step S6 for online lithium battery RUL prediction.

[0259] In this embodiment, in step S6, the phase space reconstruction matrix is obtained using the current and voltage sequence data of the unknown RUL lithium battery.

[0260] Embodiment 23:

[0261] The lithium battery rated capacity (RUL) prediction method based on phase space reconstruction and multi-space attention mechanism is described in Example 22. Further, the RUL prediction method of this example is demonstrated using lithium batteries B0005 and B0006 provided by NASA. The charging process uses a constant current of 1.5A to charge to 4.2V, and then switches to a constant voltage mode after reaching 4.2V, continuing charging until the current drops to 20mA. The discharging process uses a constant current of 2A until the battery voltage drops to a set cutoff voltage of 2.7V. The capacity degradation rate of this data shows significant differences at different stages, providing important evidence for studying the battery degradation mode and its dynamic characteristics, and helping to improve the accuracy and reliability of battery management systems in practical applications. Figure 2 The image shows the capacity degradation curves of lithium batteries B0005 and B0006.

[0262] This embodiment uses data from battery B0005 as the training set and data from battery B0006 as the test set to conduct a prediction experiment for the unknown RUL of lithium batteries. The hyperparameter settings of the gated recurrent unit network-multi-spatial attention mechanism (GRU-HAM) in this embodiment are shown in Table 1.

[0263] Table 1. Hyperparameters of Gated Recurrent Unit Network-Multi-Spatial Attention Mechanism (GRU-HAM)

[0264]

[0265] To verify the generalization ability of the method in this embodiment, current and voltage data of battery B0006 over the first 30, 50, and 70 cycles were selected. The capacity data obtained through smoothing calculations was input into the model for RUL prediction, and the model's performance under different cycle data was examined. Furthermore, it was compared with four different methods: Convolutional Neural Network (CNN), Support Vector Machine (SVM), Long Short-Term Memory Neural Network (LSTM), Gated Recurrent Unit Network, and GRU-AM attention mechanism. This embodiment uses the RUL prediction absolute error (AE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) as metrics. 2 The effectiveness of the method is measured using four evaluation indicators. The RUL prediction results for 30, 50, and 70 periods of data are as follows: Figure 3 , Figure 4 , Figure 5 As shown in the figure; the evaluation index results are shown in Table 2:

[0266] Table 2 Evaluation Index Results

[0267]

[0268] from Figures 3-5From the figure, we can clearly see that the capacity prediction trajectory of the proposed RUL prediction method is closer to the actual capacity degradation trajectory compared with the results of the other four models under the same prediction starting point. From Table 2, we can also intuitively see that the prediction error, MAE and RMSE indexes of the proposed method are lower than those of the other four RUL prediction methods. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism proposed in this embodiment can effectively capture the dynamic characteristics of the battery in different degradation stages, exhibit strong robustness and high prediction accuracy, and provide strong support for lithium battery health management and life prediction in actual applications.

Claims

1. A lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism, characterized in that, The method comprises the following steps: 1) obtaining a charging current sequence and a charging voltage sequence of each charge-discharge cycle of a lithium battery; 2) constructing a capacity phase space reconstruction model, and inputting the obtained charging current sequence and charging voltage sequence into the capacity phase space reconstruction model to obtain a capacity phase space reconstruction matrix; 3) dividing the capacity phase space reconstruction matrix into different subspaces according to a capacity degradation rate characteristic of the capacity phase space reconstruction matrix; 4) applying a multi-head attention mechanism to calculate attention scores of different subspaces, and combining the attention scores of different subspaces into a multi-space attention matrix through a fully connected layer; The multi-space attention matrix is as follows: MH = Concat(head1,...,head ir ,...,head h )W (2) HAM = σ (MH * W space + b space ) (3) where Attention is the single-head attention score, A is the query vector matrix, K is the key vector matrix, V is the key vector matrix, T is the matrix transpose, d k is the scaling factor, Concat denotes the matrix concatenation, i r is the single-head attention index, denotes the hidden state of the single-head attention i r i r ∈{1,2,...,h}, h is the number of single-head attentions, is the query vector matrix of the single-head attention i r is the key vector matrix of the single-head attention i r is the key vector matrix of the single-head attention i r σ is the activation function, W, W space are the subspace weight matrix, the multi-space weight matrix, respectively, b space is the multi-space bias; HAM is the multi-space attention matrix; MH is the spliced single-head attention matrix;​​​ 5) passing the multi-space attention matrix to a recurrent neural network GRU to construct an offline RUL prediction model with the capacity phase space reconstruction matrix as input and the remaining useful life RUL of the lithium battery as output; The offline RUL prediction model is as follows: z t = σ1(w z * h t-1 , HAM t + b z ) (4) r t = σ1(w r *[h t-1 , HAM t ]+ b r ) (5) In the formula, HAM t is the input multi-space attention matrix at time t; z t , t , h t respectively represent the update gate, the reset gate, the candidate hidden state, the output of the hidden state at time t; h t-1 is the output of the hidden state at time t-1; σ1 and σ2 are respectively a sigmoid function and a hyperbolic tangent activation function, and is a matrix element point product; w z , w r , w h are respectively the weight matrices of the update gate, the reset gate and the hidden state; b z , b r , b h are respectively the biases of the update gate, the reset gate and the hidden state. 6) collecting a charging current sequence and a charging voltage sequence of a lithium battery to be predicted, and inputting the charging current sequence and the charging voltage sequence into the capacity phase space reconstruction model to obtain a capacity phase space reconstruction matrix of the lithium battery to be predicted; 7) inputting the capacity phase space reconstruction matrix of the lithium battery to be predicted into the offline RUL prediction model to predict the remaining useful life RUL of the lithium battery to be predicted.

2. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 1, characterized in that, In step 2), the steps of obtaining the capacity phase space reconstruction matrix are as follows: 2.1) performing smoothing processing on the data input into the capacity phase space reconstruction model to obtain processed data; 2.2) calculating a battery capacity based on the processed data; 2.3) determining an optimal embedding dimension and a time delay by using a C-C algorithm based on the battery capacity, and constructing a capacity phase space reconstruction matrix.

3. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 2, characterized in that, The smoothing processing adopts a sliding mean filter, as shown below: In the formula, i is the charge and discharge cycle index; t represents the time; p is the data collection number; and L is the sliding window size. respectively represent the current and voltage data at time t in the i th cycle after smoothing processing. i,t , V i,t respectively represent the current and voltage data at time t in the i th cycle after smoothing processing.

4. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 2, characterized in that, The battery capacity is as follows: In the formula, i is the charge and discharge cycle index, n is the total cycle number; t represents the time; p is the data collection number, and e is the end time; respectively represent the current and voltage data at time t in the i th cycle after smoothing processing; V nom is the nominal voltage; Q i represents the charge capacity of the i th cycle.

5. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 2, characterized in that, The battery capacity also adopts a normalization processing to standardize the data, and the calculation formula of the normalization processing is as follows: In the formula, X * is the normalized data, X is the data to be normalized, X max , X min are the maximum value and minimum value in the sequence X, respectively.

6. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 2, characterized in that, In step 2.3), the steps of determining the optimal embedding dimension and the time delay by using the C-C algorithm are as follows: 2.3.1) calculating a cross-correlation integral based on the battery capacity, as shown below: where C(m, n, r, τ) is the cross-correlation integral; i, j are both charge and discharge cycle indexes, n is the total cycle number; Q i , Q j represent the charge capacity of the i, j th cycle, respectively; m represents the embedding dimension, M is the number of embedding points in the m-dimensional space, and M = n - (m - 1)τ, τ represents the time delay, and r represents the spatial distance; Wherein, the piecewise function F is as follows: 2.3.2) calculating a nonlinear feature function based on the cross-correlation integral and a time delay average strategy, as shown below: where S(m, n, r, t) is a non-linear characteristic function, S denotes the time delay index, and denote the cross-correlation integrals; 2.3.3) converting the nonlinear feature function S(m, n, r, τ) into a minimum position of a change rate of an arbitrary spatial distance r, and deriving an optimal time delay τ; The minimum position of the change rate of the arbitrary spatial distance r is as follows: ΔS(m,τ)=max{S(m,r j ,τ)}-min{S(m,r j ,t)} (14) where ΔS(m,τ) represents the position of the minimum rate of change of the arbitrary spatial distance r, S(m,r j ,τ) represents a nonlinear characteristic function of the arbitrary spatial distance r when n is infinite. 2.3.4) solving an optimal embedding dimension based on a time delay window, as shown below: τ window = (m - 1)τ (15) In the formula, τ window is a time delay window; is an average of the minimum position of the rate of change of the arbitrary spatial distance r; is an average of the nonlinear characteristic function of the arbitrary spatial distance r.

7. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 2, characterized in that, The capacity phase space reconstruction matrix is as follows: where i is the charge-discharge cycle index, n is the total cycle number, m represents the embedding dimension, Q i represents the charge capacity of the i-th cycle; Q R , Q No respectively represent the regenerated incremental capacity and the non-regenerated incremental capacity.

8. The lithium battery RUL prediction method based on phase space reconstruction and multi-space attention mechanism according to claim 7, characterized in that, said regenerative incremental capacity Q R and no regenerative incremental capacity Q No as follows: In the formula, q is an upper limit of a regenerated incremental capacity.

Images