Lithium battery state of health prediction method and device based on multi-scale analysis, equipment and storage medium

Abstract

This invention discloses a method, apparatus, device, and storage medium for predicting the health status of lithium batteries based on multi-scale analysis. The method includes: acquiring multi-dimensional time-series data of the lithium battery under driving and laboratory conditions; determining the cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value based on the multi-dimensional time-series data; obtaining dynamic time-series features and static characteristic parameters; performing multi-scale partitioning based on the dynamic time-series features to obtain an updated time series; inputting the updated time series into a selective representation space module for prediction to obtain multiple health status estimates; and fusing the multiple health status estimates based on the static characteristic parameters to obtain a target health status value. This invention introduces multi-dimensional time-series data of lithium batteries, improves prediction performance through physical constraints, and effectively extracts implicit information from the charging and discharging process through multi-scale analysis, thereby improving the accuracy and robustness of the target health status value of the lithium battery.

Description

Technical Field

[0001] This invention relates to the field of lithium battery health status prediction technology, and in particular to a method, apparatus, device and storage medium for predicting the health status of lithium batteries based on multi-scale analysis. Background Technology

[0002] Lithium-ion batteries, as the most widely used electrochemical energy storage devices, play an irreplaceable role in fields such as transportation electrification and smart grids. Accurately predicting the health status of lithium-ion batteries is a crucial prerequisite for ensuring the safe and reliable operation of equipment and optimizing battery management system strategies. However, lithium-ion batteries are inherently highly complex dynamic electrochemical systems with nonlinear behavior and intricate internal mechanisms, posing significant challenges to health status prediction.

[0003] Battery aging is the result of multiple physicochemical processes. Physically, thermal stress cycling-induced volume expansion or contraction of electrode materials and mechanical stress-induced particle cracking and contact failure both accelerate performance degradation. Chemically, lithium inventory loss and active material dissolution are side effects that contribute to battery degradation, forming a difficult-to-decouple degradation mechanism. Battery degradation not only occurs during charge-discharge cycles but also naturally during storage. Under different operating conditions, the contribution weights of the aforementioned dominant mechanisms change dynamically, and the degradation trajectory exhibits non-linear characteristics. Single-scale analysis methods struggle to simultaneously capture rapidly changing transient features and slowly evolving long-term trends. Summary of the Invention

[0004] Therefore, it is necessary to propose a method, device, equipment, and storage medium for predicting the health status of lithium batteries based on multi-scale analysis to address the above problems.

[0005] A method for predicting the health status of lithium batteries based on multi-scale analysis, the method comprising:

[0006] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0007] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0008] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0009] In one embodiment,

[0010] The determination of the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current is achieved by the following expression:

[0011]

[0012]

[0013] in, express The cumulative ampere-hour integral value at each moment, Indicates the first The first current at each sampling time Indicates the length of the sliding time window. Indicates the sampling time interval; This represents the heat production power per unit time. Indicates time The first current, This represents the equivalent resistance inside the battery, taken as... As a characteristic of heat-generating factors;

[0014] The determination of the current-temperature interaction feature value based on the first current and the first temperature is achieved by the following expression:

[0015]

[0016] in, Represents the current-temperature interaction characteristic value. Indicates time The first temperature.

[0017] In one embodiment, obtaining the static feature parameters based on the second and third original multidimensional time series data includes:

[0018] The instantaneous voltage drop of the lithium battery is obtained by using the second voltage at adjacent time points; the total voltage drop of the lithium battery is obtained by using the second voltage during the pulse discharge interval.

[0019] At a specific temperature, based on Ohm's law, the ohmic resistance and polarization resistance are calculated using the instantaneous voltage drop, total voltage drop, and preset instantaneous current during the pulse discharge of a lithium battery.

[0020] Based on the Nyquist plot, the intersection of the high-frequency region and the real axis is extracted using the real and imaginary parts of the complex impedance at different frequencies as the ohmic impedance, and the diameter of the semicircular arc in the mid-frequency region is extracted using the real and imaginary parts of the complex impedance at different frequencies as the charge transfer impedance.

[0021] The ohmic resistance, the polarization resistance, the ohmic impedance, and the charge transfer impedance constitute static characteristic parameters.

[0022] In one embodiment, the ohmic resistor and the polarization resistor are implemented by the following expression:

[0023]

[0024]

[0025] in, This refers to the instantaneous voltage drop of the lithium battery. This refers to the total voltage drop of the lithium battery. This is the preset instantaneous current for the lithium battery; Indicates ohmic resistance; This indicates the polarization resistance.

[0026] In one embodiment, the normalization process of the dynamic time series features to obtain normalized dynamic time series features includes:

[0027]

[0028]

[0029] in, , They represent the first The mean and standard deviation of each channel; The dynamic temporal characteristics of the c-th channel at time t; The length of the time series; Represents the numerical stability constant;

[0030]

[0031] in, Normalized dynamic time series characteristics; The dynamic time series feature matrix is ​​composed of dynamic time series features. ,in Indicates batch size, Indicates the length of the time series. Indicates the number of feature channels; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

[0032] In one embodiment, the step of performing multi-scale partitioning of the normalized dynamic time series features to obtain the updated time series includes:

[0033] Normalized dynamic time series characteristics Downsampling is performed, and average pooling is used to obtain... Time series at various scales It can be obtained through the following formula:

[0034]

[0035]

[0036] in, For the first Time series at various scales; For the first Time series at various scales; The base number for downsampling; For average pooling layers; For pooling cores; The pooling step size; This is a time series at scale 0; For dimension rearrangement operations; This is a downsampling level index;

[0037] For the Time series at various scales Information exchange is performed through MLP to obtain updated time series data. ,as follows;

[0038]

[0039]

[0040] in, For the first Time series at various scales Updated time series; For random deactivation layers; It is a multilayer perceptron; To update the time series.

[0041] In one embodiment, fusing multiple health status estimates based on static feature parameters to obtain the target health status value includes:

[0042]

[0043] in, Represents the fused health status estimate; i represents the i-th scale. k represents the number of scales obtained from downsampling; This represents the weight of the i-th scale; This represents the single-scale health status estimate at the i-th scale;

[0044] The target health state value is obtained by inverse normalizing the fused health state estimate through the RevIN layer, and is calculated using the following formula:

[0045]

[0046] in, Indicates the target health status value; This represents the integrated health status estimate; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

[0047] A lithium battery health status prediction device based on multi-scale analysis, the device comprising:

[0048] The acquisition module is used to acquire the first raw multidimensional time-series data of the lithium battery under driving conditions, including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0049] The determination module is used to determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, the first current, the first temperature, the first ampere-hour, the cumulative ampere-hour integral value, the heat generation power per unit time, and the current-temperature interaction characteristic value constitute dynamic time-series characteristics; and obtain static characteristic parameters based on the second original multidimensional time-series data and the third original multidimensional time-series data.

[0050] The fusion module is used to normalize the dynamic time series features to obtain normalized dynamic time series features; to perform multi-scale partitioning on the normalized dynamic time series features to obtain an updated time series; to input the updated time series into the selective representation space module for prediction to obtain multiple health state estimates; and to fuse the multiple health state estimates based on static feature parameters to obtain the target health state value.

[0051] A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps:

[0052] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0053] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0054] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0055] A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps:

[0056] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0057] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0058] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0059] This invention introduces multi-dimensional time-series data of lithium batteries and improves prediction performance through physical constraints. Through multi-scale analysis methods, it can effectively extract implicit information of the charging and discharging process, thereby improving the accuracy and robustness of the target health state value of lithium batteries. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] in:

[0062] Figure 1 This is an application environment diagram of a lithium battery health status prediction method based on multi-scale analysis in one embodiment;

[0063] Figure 2This is a flowchart of a lithium battery health status prediction method based on multi-scale analysis in one embodiment;

[0064] Figure 3 This is a structural block diagram of a lithium battery health status prediction device based on multi-scale analysis in one embodiment;

[0065] Figure 4 This is a structural block diagram of a computer device in one embodiment. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] To address the technical problems in the background art, this application provides a method for predicting the health status of lithium batteries based on multi-scale analysis.

[0068] Figure 1 This is an application environment diagram of a lithium battery health state prediction method based on multi-scale analysis in one embodiment. (Refer to...) Figure 1This multi-scale analysis-based lithium battery health state prediction method is applied to a multi-scale analysis-based lithium battery health state prediction system. The system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal, specifically a mobile phone, tablet computer, laptop computer, or similar device. The server 120 can be a standalone server or a server cluster consisting of multiple servers. The terminal 110 is used to acquire first raw multi-dimensional time-series data of the lithium battery under driving conditions, including first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; second raw multi-dimensional time-series data under laboratory conditions, obtained by testing the lithium battery using a five-pulse discharge HPPC test, including a second voltage; and third raw multi-dimensional time-series data under laboratory conditions, obtained by testing the lithium battery using an EIS test, including the real part and imaginary part of the complex impedance. Server 120 is used to determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction feature value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction feature value constitute dynamic time series features; obtain static feature parameters based on the second original multidimensional time series data and the third original multidimensional time series data; normalize the dynamic time series features to obtain normalized dynamic time series features; perform multi-scale partitioning on the normalized dynamic time series features to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health state estimates; and fuse the multiple health state estimates based on the static feature parameters to obtain the target health state value.

[0069] like Figure 2 As shown, in one embodiment, a method for predicting the health status of lithium batteries based on multi-scale analysis is provided. This method can be applied to both terminals and servers; this embodiment illustrates its application to terminals. Figure 2 As shown, the lithium battery health status prediction method based on multi-scale analysis specifically includes the following steps:

[0070] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0071] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0072] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0073] In one embodiment,

[0074] The determination of the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current is achieved by the following expression:

[0075]

[0076]

[0077] in, express The cumulative ampere-hour integral value at each moment, Indicates the first The first current at each sampling time Indicates the length of the sliding time window. Indicates the sampling time interval; This represents the heat production power per unit time. Indicates time The first current, This represents the equivalent resistance inside the battery, taken as... As a characteristic of heat-generating factors;

[0078] The determination of the current-temperature interaction feature value based on the first current and the first temperature is achieved by the following expression:

[0079]

[0080] in, Represents the current-temperature interaction characteristic value. Indicates time The first temperature.

[0081] In one embodiment, obtaining the static feature parameters based on the second and third original multidimensional time series data includes:

[0082] The instantaneous voltage drop of the lithium battery is obtained by using the second voltage at adjacent time points. The total voltage drop of the lithium battery is obtained by using the second voltage in the pulse discharge range. ;

[0083] At a specific temperature, according to Ohm's law, the instantaneous voltage drop during the pulse discharge of a lithium battery... Total voltage drop and preset instantaneous current Calculate ohmic resistance and polarization resistance ;

[0084] Based on the Nyquist plot, the real part of the complex impedance at different frequencies is used. and the imaginary part of the complex impedance The intersection of the high-frequency region and the real axis is extracted as the ohmic impedance. Using the real part of the complex impedance at different frequencies and the imaginary part of the complex impedance Extract the diameter of the semicircular arc in the mid-frequency region as the charge transfer impedance. ;

[0085] The ohm resistor The polarization resistor The ohmic impedance and the charge transfer impedance These constitute static characteristic parameters.

[0086] In one embodiment, the ohmic resistor and polarization resistance This can be achieved using the following expression:

[0087]

[0088]

[0089] in, This refers to the instantaneous voltage drop of the lithium battery. This refers to the total voltage drop of the lithium battery. This is the preset instantaneous current for the lithium battery; Indicates ohmic resistance; This indicates the polarization resistance.

[0090] In one embodiment, the dynamic temporal features Normalization processing yields the following normalized dynamic time-series features:

[0091]

[0092]

[0093] in, , They represent the first The mean and standard deviation of each channel; The dynamic temporal characteristics of the c-th channel at time t; The length of the time series; Represents the numerical stability constant;

[0094]

[0095] in, Normalized dynamic time series characteristics; The dynamic time series feature matrix is ​​composed of dynamic time series features. ,in Indicates batch size, Indicates the length of the time series. Indicates the number of feature channels; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

[0096] In one embodiment, the step of performing multi-scale partitioning of the normalized dynamic time series features to obtain the updated time series includes:

[0097] Normalized dynamic time series characteristics Downsampling is performed, and average pooling is used to obtain... Time series at various scales It can be obtained through the following formula:

[0098]

[0099]

[0100] in, For the first Time series at various scales; For the first Time series at various scales; The base number for downsampling; For average pooling layers; For pooling cores; The pooling step size; This is a time series at scale 0; For dimension rearrangement operations; This is a downsampling level index;

[0101] For the Time series at various scales Information exchange is performed through MLP to obtain updated time series data. ,as follows;

[0102]

[0103]

[0104] in, For the first Time series at various scales Updated time series; For random deactivation layers; It is a multilayer perceptron; To update the time series.

[0105] In one embodiment, inputting the updated time series data into a selective representation space module for prediction to obtain multiple health state estimates includes:

[0106] For updating time series A replication-filling strategy is used to extend the time series boundaries, and the filled time series is obtained after the filling operation. :

[0107]

[0108] in, Indicates filling time series; Indicates the copy-fill strategy; This indicates an update to the time series.

[0109] After sequence filling, local continuous segments of the filled time series are extracted using a standard sliding window, and the dimensions of these local continuous segments are rearranged to obtain the original view of the sequence. :

[0110]

[0111] in, Represents the original view of the sequence; This indicates a dimension rearrangement operation; This indicates a sliding window operation; size represents the length of the sliding window, set to... ;step represents the step size of the sliding window, set to S.

[0112] Local continuous segments of the filled time series are extracted using a dense sliding window with a step size of 1, and the dimensions of these local continuous segments are rearranged to construct a reconstructed view. :

[0113]

[0114] in, Indicates a refactored view; This indicates a dimension rearrangement operation; This indicates a sliding window operation; size represents the length of the sliding window, set to... ;step represents the step size of the sliding window, which is set to 1.

[0115] The selective block segmentation function is implemented through an MLP-based weighted scorer. In practice, this weighted scorer can calculate the attention weight of each patch in the reconstructed view, sort them according to the attention weight, and obtain a sorted subset. .

[0116]

[0117] in, Represents a sorted subset; The weighted scorer is represented by MLP; MLP represents a multilayer perceptron. This indicates a refactored view.

[0118] By performing aggregation operations, the N patches with the highest attention weights are extracted from the reconstructed view, and a sparse candidate set is constructed. :

[0119]

[0120] in, Represents a sparse candidate set; Indicates aggregation operation; Indicates a refactored view; Used to return the index positions of the N largest values ​​in the attention score matrix; This represents a sorted subset.

[0121] Patches in the sparse candidate set are weighted according to attention weights and then re-ranked using an MLP-based scorer. Reorder the subsets. The resulting reordered subset B is:

[0122]

[0123]

[0124] in, Represents a weighted sparse candidate set; Indicates attention weight; B represents a sparse candidate set; B represents a reordered subset. represents a reordering scorer; MLP represents a multilayer perceptron.

[0125] Reorder the reordered subset in descending order to obtain its descending index. Based on the descending index, the sequence is rearranged through aggregation operations to obtain a reconstructed sequence view. :

[0126]

[0127]

[0128] in, Indicates a descending index; B represents a descending order reordering operation; B represents a reordered subset. Represents a sequence-reconstructed view; This indicates a dimension rearrangement operation; This represents a weighted sparse candidate set.

[0129] The dual-view adaptive fusion layer dynamically balances the contribution weights of the original view and the reconstructed view through learnable parameters to generate the optimal fusion representation.

[0130] Calculate the linear projection of the original sequence view and the reconstructed sequence view, linearly mapping the data dimensions to the model dimensions:

[0131]

[0132]

[0133] in, This indicates that the original view is being mapped. Represents the original view of the sequence; This indicates a mapping to reconstruct the view; Represents a sequence-reconstructed view; This represents a linear mapping operation.

[0134] fusion weights are generated using the Sigmoid function. :

[0135]

[0136] in, Indicates the fusion weights; These are learnable parameters.

[0137] Finally, a weighted fusion operation is performed on the original mapped view and the reconstructed mapped view to obtain the fused output. :

[0138]

[0139] in, Indicates fused output; Indicates the fusion weights; This indicates that the original view is being mapped. This indicates a mapping to reconstruct the view.

[0140] The location embedding layer marks the spatiotemporal position of the selected patch in the original sequence using a location information matrix P, while preserving the temporal logic.

[0141] For fusion output Add location information and regularize it using the Dropout layer:

[0142]

[0143] in, This indicates that location information is embedded in the output; Indicates a randomly deactivated layer; This indicates the fused output; P represents the location information matrix.

[0144] The prediction output layer embeds location information into the output and maps it to the prediction length, resulting in a single-scale health status estimate. :

[0145]

[0146] in, This represents a single-scale health status estimate; This represents a multilayer perceptron; Indicates the flattening operation; This indicates that location information is embedded in the output.

[0147] The prediction results at different scales are weighted and the results are output through the scoring module, specifically including:

[0148] Based on static features, the single-scale health status estimates are fused to obtain a fused health status estimate:

[0149] static feature parameters Mapped to weight distribution function The weighted distribution function assigns weights to the health status estimates at different scales, and is obtained through the following formula:

[0150]

[0151] in, Let be the weight distribution function. The dimension corresponds to the number of scales in multi-scale prediction; This represents the normalized exponential function; This represents a multilayer perceptron; Represents static characteristic parameters.

[0152] In one embodiment, the estimation of multiple health states based on static feature parameters The target health status value is obtained by fusion. include:

[0153]

[0154] in, Represents the fused health status estimate; i represents the i-th scale. k represents the number of scales obtained from downsampling; This represents the weight of the i-th scale; This represents the single-scale health status estimate at the i-th scale;

[0155] The target health state value is obtained by inverse normalizing the fused health state estimate through the RevIN layer, and is calculated using the following formula:

[0156]

[0157] in, Indicates the target health status value; This represents the integrated health status estimate; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

[0158] This invention also provides a lithium battery health status prediction device based on multi-scale analysis, such as... Figure 3 As shown, the device includes:

[0159] The acquisition module 10 is used to acquire the first raw multidimensional time-series data of the lithium battery under driving conditions, including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0160] The determination module 20 is used to determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, the first current, the first temperature, the first ampere-hour, the cumulative ampere-hour integral value, the heat generation power per unit time, and the current-temperature interaction characteristic value constitute dynamic time-series characteristics; and obtain static characteristic parameters based on the second original multidimensional time-series data and the third original multidimensional time-series data.

[0161] The fusion module 30 is used to normalize the dynamic time series features to obtain normalized dynamic time series features; to perform multi-scale division on the normalized dynamic time series features to obtain an updated time series; to input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; and to fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0162] This invention introduces multi-dimensional time-series data of lithium batteries and improves prediction performance through physical constraints. Through multi-scale analysis methods, it can effectively extract implicit information of the charging and discharging process, thereby improving the accuracy and robustness of the target health state value of lithium batteries.

[0163] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the following steps:

[0164] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0165] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0166] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0167] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps:

[0168] S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance.

[0169] S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data;

[0170] S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

[0171] Figure 4 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 4 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement a lithium battery health state prediction method based on multi-scale analysis. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to implement the lithium battery health state prediction method based on multi-scale analysis. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0172] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for lithium battery state of health prediction based on multi-scale analysis, characterized in that, The method includes: S1: Obtain the first raw multidimensional time-series data of the lithium battery under driving conditions: including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance. S2: Determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, first current, first temperature, first ampere-hour, cumulative ampere-hour integral value, heat generation power per unit time, and current-temperature interaction characteristic value constitute dynamic time series characteristics; obtain static characteristic parameters based on the second original multidimensional time series data and the third original multidimensional time series data; S3: Normalize the dynamic time series features to obtain normalized dynamic time series features; divide the normalized dynamic time series features into multiple scales to obtain an updated time series; input the updated time series into the selective representation space module for prediction to obtain multiple health status estimates; fuse the multiple health status estimates based on static feature parameters to obtain the target health status value.

2. The lithium battery health status prediction method based on multi-scale analysis according to claim 1, characterized in that, The determination of the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current is achieved by the following expression: wherein, represents the accumulated ampere-hour integral value at the time point, represents the first current at the i-th sampling time point, represents the first current at the i-th sampling time point, represents the sliding time window length, represents the sampling time interval; represents the heat generation power per unit time, represents the first current at the time point, represents the first current at the time point, represents the internal equivalent resistance of the battery, and takes as the heat generation factor characteristic; The determination of the current-temperature interaction feature value based on the first current and the first temperature is achieved by the following expression: in, Represents the current-temperature interaction characteristic value. Indicates time The first temperature.

3. The lithium battery health status prediction method based on multi-scale analysis according to claim 1, characterized in that, The static feature parameters obtained based on the second and third original multidimensional time series data include: The instantaneous voltage drop of the lithium battery is obtained by using the second voltage at adjacent time points; the total voltage drop of the lithium battery is obtained by using the second voltage during the pulse discharge interval. At a specific temperature, based on Ohm's law, the ohmic resistance and polarization resistance are calculated using the instantaneous voltage drop, total voltage drop, and preset instantaneous current during the pulse discharge of a lithium battery. Based on the Nyquist plot, the intersection of the high-frequency region and the real axis is extracted using the real and imaginary parts of the complex impedance at different frequencies as the ohmic impedance, and the diameter of the semicircular arc in the mid-frequency region is extracted using the real and imaginary parts of the complex impedance at different frequencies as the charge transfer impedance. The ohmic resistance, the polarization resistance, the ohmic impedance, and the charge transfer impedance constitute static characteristic parameters.

4. The lithium battery health status prediction method based on multi-scale analysis according to claim 3, characterized in that, The ohmic resistance and polarization resistance are achieved by the following expression: in, This refers to the instantaneous voltage drop of the lithium battery. This refers to the total voltage drop of the lithium battery. This is the preset instantaneous current for the lithium battery; Indicates ohmic resistance; This indicates the polarization resistance.

5. The lithium battery health status prediction method based on multi-scale analysis according to claim 1, characterized in that, The normalization process for the dynamic time series features to obtain normalized dynamic time series features includes: in, , They represent the first The mean and standard deviation of each channel; The dynamic temporal characteristics of the c-th channel at time t; The length of the time series; Represents the numerical stability constant; in, Normalized dynamic time series characteristics; The dynamic time series feature matrix is ​​composed of dynamic time series features. ,in Indicates batch size, Indicates the length of the time series. Indicates the number of feature channels; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

6. The lithium battery health status prediction method based on multi-scale analysis according to claim 1, characterized in that, The step of performing multi-scale partitioning on the normalized dynamic time series features to obtain the updated time series includes: Normalized dynamic time series characteristics Downsampling is performed, and average pooling is used to obtain... Time series at various scales It can be obtained through the following formula: in, For the first Time series at various scales; For the first Time series at various scales; The base number for downsampling; For average pooling layers; For pooling cores; The pooling step size; This is a time series at scale 0; For dimension rearrangement operations; This is a downsampling level index; For the Time series at various scales Information exchange is performed through MLP to obtain updated time series data. ,as follows; in, For the first Time series at various scales Updated time series; For random deactivation layers; It is a multilayer perceptron; To update the time series.

7. The lithium battery health status prediction method based on multi-scale analysis according to claim 1, characterized in that, The process of fusing multiple health status estimates based on static feature parameters to obtain the target health status value includes: in, Represents the fused health status estimate; i represents the i-th scale. k represents the number of scales obtained from downsampling; This represents the weight of the i-th scale; This represents the single-scale health status estimate at the i-th scale; The target health state value is obtained by inverse normalizing the fused health state estimate through the RevIN layer, and is calculated using the following formula: in, Indicates the target health status value; This represents the integrated health status estimate; The average of multiple channels The mean matrix formed; It is a standard deviation matrix composed of the standard deviations of multiple channels.

8. A lithium battery health status prediction device based on multi-scale analysis, characterized in that, The device includes: The acquisition module is used to acquire the first raw multidimensional time-series data of the lithium battery under driving conditions, including the first voltage, first current, first temperature, and first ampere-hour during driving cycle discharge; the second raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through a five-pulse discharge HPPC test, including the second voltage; and the third raw multidimensional time-series data under laboratory conditions, which is obtained by testing the lithium battery through an EIS test, including the real part and imaginary part of the complex impedance. The determination module is used to determine the cumulative ampere-hour integral value and the heat generation power per unit time based on the first current; determine the current-temperature interaction characteristic value based on the first current and the first temperature; the first voltage, the first current, the first temperature, the first ampere-hour, the cumulative ampere-hour integral value, the heat generation power per unit time, and the current-temperature interaction characteristic value constitute dynamic time-series characteristics; and obtain static characteristic parameters based on the second original multidimensional time-series data and the third original multidimensional time-series data. The fusion module is used to normalize the dynamic time series features to obtain normalized dynamic time series features; to perform multi-scale partitioning on the normalized dynamic time series features to obtain an updated time series; to input the updated time series into the selective representation space module for prediction to obtain multiple health state estimates; and to fuse the multiple health state estimates based on static feature parameters to obtain the target health state value.

9. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 7.

Images