Lithium battery SOC / SOH joint estimation method based on parameter updating mechanism model and LSTM algorithm

Abstract

The invention belongs to the technical field of voltage prediction, and relates to a lithium battery SOC / SOH joint estimation method based on a parameter updating mechanism model and an LSTM algorithm, and the method comprises the steps: estimating the SOC of a lithium battery through employing a Kalman filtering AEKF algorithm; estimating the SOH (state of health) of the lithium battery by using the LSTM neural network; updating the input of the LSTM neural network through the estimated state of charge SOC, correcting the parameters of Kalman filtering AEKF through the estimated state of health SOH, and constructing a joint estimation model; and calculating a loss function of the joint estimation model, and training the joint estimation model to realize joint estimation of the state of charge (SOC) and the state of health (SOH). According to the method, the charge state and the health state of the battery are comprehensively considered and are coupled, so that the evaluation precision of the health state of the battery is effectively improved, the full-life-cycle management of the battery is optimized, and the energy utilization efficiency and the safety of the battery are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of voltage prediction, and in particular to a lithium battery SOC / SOH joint estimation method based on a parameter update mechanism model and an LSTM algorithm. Background Art

[0002] Lithium-ion batteries are widely used in electric vehicles and electrochemical energy storage systems due to their high energy, power density, and long cycle life. The state of charge (SOC) and state of health (SOH) of lithium-ion batteries are crucial for the safe and efficient operation of battery systems. SOC provides information on the battery's current remaining charge, which is critical for energy management in applications such as electric vehicles and energy storage systems. SOH reflects the battery's aging and performance degradation, helping to predict when maintenance or battery replacement is needed, thereby improving system reliability.

[0003] Batteries age over time, causing changes in their capacity parameters. Traditional SOC estimation methods typically assume that the battery capacity remains constant, ignoring the impact of battery aging, which can lead to inaccurate SOC estimation results.

[0004] In existing technologies, SOC and SOH are typically estimated independently, ignoring their mutual influence. Traditional SOC estimation methods assume a constant battery capacity, but as the battery ages, capacity decay and increased internal resistance can affect SOC estimation accuracy. Similarly, SOH estimation often fails to fully account for dynamic changes in SOC, leading to errors in state-of-health predictions. A battery's SOC and SOH are tightly coupled throughout its lifecycle. Changes in SOC reflect some characteristics of its state of health, while changes in SOH directly affect SOC estimation.

[0005] In practical applications, SOC and SOH estimation often interfere with each other. As batteries age, the errors caused by this interference are amplified, causing the accuracy of both estimates to decline. This not only affects battery efficiency but can also shorten their service life, increasing system operating costs and posing safety risks. Therefore, improving the accuracy of SOC and SOH estimation for aging batteries and reducing the mutual interference between the two are key issues that need to be addressed in the current battery management field. Summary of the Invention

[0006] In response to the technical problems raised above, a method for joint estimation of SOC / SOH of lithium batteries based on a parameter update mechanism model and an LSTM algorithm is provided. The present invention extracts a battery life health factor applicable to online scenarios and completes the SOH estimation of lithium-ion batteries based on a long short-term memory neural network algorithm. The SOH estimation algorithm is integrated into an AEKF-based SOC estimation algorithm with online model parameter updates, and a multi-time scale update strategy for SOC and SOH is formulated. The present invention promptly corrects the capacity parameters in the SOC estimation in the case of battery aging, ensuring its estimation accuracy under different working conditions.

[0007] The technical means adopted in the present invention are as follows:

[0008] A lithium battery SOC / SOH joint estimation method based on a parameter update mechanism model and an LSTM algorithm includes:

[0009] Use Kalman filter AEKF algorithm to estimate the state of charge SOC of lithium battery;

[0010] Use LSTM neural network to estimate the health status SOH of lithium batteries;

[0011] The input of the LSTM neural network is updated by the estimated state of charge (SOC), and the parameters of the Kalman filter (AEKF) are corrected by the estimated state of health (SOH) to build a joint estimation model.

[0012] The loss function of the joint estimation model is calculated and the joint estimation model is trained to achieve joint estimation of the state of charge (SOC) and the state of health (SOH).

[0013] Furthermore, when estimating the state of charge (SOC), a nonlinear model is constructed using the electrochemical model SP+ based on lithium-ion batteries, the state of charge (SOC) of the lithium battery is used as a system state variable, the current is used as input, and the terminal voltage of the lithium battery is output;

[0014] The Kalman filter (AEKF) algorithm runs once per second in the full-life state-of-charge (SOC) estimation algorithm and continuously adjusts parameters. After each parameter adjustment, the parameter value of the state-of-charge (SOC) estimation is updated to adapt to the changes in the lithium-ion battery over time.

[0015] Furthermore, the LSTM neural network introduces memory units and a gating mechanism to estimate the state of health (SOH) of the lithium battery;

[0016] The gating mechanism includes an input gate, a forget gate, and an output gate. The mathematical expression of the forget gate is:

[0017]

[0018] Among them, ft represents the forget gate, σ represents the Sigmoid activation function, W f represents the forget gate weight matrix, h t-1 Represents the hidden state of the previous moment, x t represents the input at time t, b f Represents the bias vector matrix.

[0019] The mathematical expression of the input gate is:

[0020]

[0021] Among them, i t represents the input gate, represents the candidate memory cell state, W C represents the weight matrix, b i 、b C Represents the bias vector matrix.

[0022] The mathematical expression of the memory unit is:

[0023]

[0024] Among them, C t Represents the memory cell state of LSTM at time t.

[0025] The mathematical expression of the output gate is:

[0026] o t =σ(W o ·[h t-1 ,x t ]+b o )h t =o t tanh(C t )

[0027] Among them, t represents the output gate, W o represents the output gate weight matrix, b o represents the bias vector matrix, h t represents the hidden state at time t.

[0028] Furthermore, the joint estimation model specifically includes:

[0029] The input data includes the charge and discharge time series data of the lithium battery X = (x1, x2, ... x t ); after normalization, the eigenvalues ​​are scaled to the same scale and divided into time windows of fixed length; the LSTM algorithm is used to capture the long-term dependencies in the lithium battery charge and discharge data. The hidden state update formula of the LSTM algorithm is:

[0030] h t =LSTM(x t ,h t-1 )

[0031] Among them, x t is the input at time t, h t-1 It is the hidden state of the previous moment, adopts a multi-layer LSTM structure, and adds a Dropout layer to prevent overfitting.

[0032] The output of the LSTM algorithm is passed through a shared fully connected layer to extract features:

[0033] z=ReLU(W h ·h T +b h )

[0034] Among them, z represents the output feature vector, h T represents the hidden state of LSTM, represents the weight of the shared layer, and b h Represents the bias of the shared layer; the output of the shared layer is mapped to the estimated values ​​of SOC and SOH through two independent fully connected layers:

[0035] SOC=σ(W soc z+b soc )

[0036] SOH=σ(W soh z+b soh )

[0037] Among them, W soc represents the weight of the SOC branch, b soc represents the bias of the SOC branch; W soh represents the weight of the SOH branch, b soh represents the bias of the SOH branch, and σ is the Sigmoid activation function.

[0038] Furthermore, the loss function of the joint estimation model is calculated:

[0039] The loss function L of the joint estimation model is the weighted sum of SOC loss and SOH loss:

[0040] L=α·L soc +(1-α)·L soh

[0041] Among them, L soc represents the mean square error of SOC, L soh represents the mean square error of SOH, α is the weight parameter;

[0042] The Adam optimizer is used for model training, and the early stopping strategy is used to prevent overfitting.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] The present invention provides a combined SOC / SOH estimation method for lithium batteries based on a parameter update mechanism model and an LSTM algorithm. This combined estimation method, powered by an LSTM neural network, enables real-time correction of battery capacity parameters, resolving the issue of traditional SOC estimation that ignores the impact of battery aging. By dynamically adjusting SOC estimation under varying operating conditions, the method ensures accurate SOC estimation results even with battery aging, thereby improving the reliability and accuracy of the battery management system under various operating conditions.

[0045] The proposed method for lithium battery SOC / SOH joint estimation, based on a parameter update mechanism model and an LSTM algorithm, fully considers the mutual influence between SOC and SOH. The proposed joint estimation framework can accurately estimate SOC on a short time scale and track SOH trends on a long time scale, avoiding the mutual interference between SOC and SOH estimation in existing technologies. This method, which comprehensively considers the coupling effect of the two, effectively improves the accuracy of battery health assessment, optimizes the battery's full life cycle management, and enhances the battery's energy utilization efficiency and safety.

[0046] Based on the above reasons, the present invention can be widely promoted in fields such as voltage prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0048] Figure 1 This is a flow chart of the lithium battery SOC / SOH joint estimation method based on the parameter update mechanism model and LSTM algorithm in the present invention.

[0049] Figure 2 This is the structural diagram of the AEKF algorithm in the present invention.

[0050] Figure 3 Schematic diagram of SOH estimation based on the LSTM neural network model in the present invention.

[0051] Figure 4 This is a framework diagram of the joint estimation model in the present invention.

[0052] Figure 5 This is the battery terminal voltage curve under NEDC operating conditions in different aging states in the present invention.

[0053] Figure 6 This is a comparison chart of SOC estimation results of different algorithms in different aging states in the present invention.

[0054] Figure 7 This is a comparison chart of the SOH estimation results and the measured values ​​in the present invention. DETAILED DESCRIPTION

[0055] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is in no way intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0057] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0058] Unless otherwise specified, the relative arrangement of the parts and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be clear that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to actual proportional relationships. The technology, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be considered as a part of the specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of the exemplary embodiments can have different values. It should be noted that similar numbers and letters represent similar items in the following drawings, and therefore, once an item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0059] In the description of the present invention, it should be understood that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "horizontal, vertical, vertical, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention: the directional words "inside and outside" refer to the inside and outside relative to the outline of each component itself.

[0060] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used herein to describe the spatial positional relationship of a device or feature to other devices or features as shown in the figures. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figures. For example, if the device in the drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below their position devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0061] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of the present invention.

[0062] like Figure 1 As shown, the present invention provides a lithium battery SOC / SOH joint estimation method based on a parameter update mechanism model and an LSTM algorithm, comprising:

[0063] The Kalman filter (AEKF) algorithm is used to estimate the state of charge (SOC) of the lithium battery.

[0064] In a specific implementation, as a preferred embodiment of the present invention, when estimating the state of charge (SOC), a nonlinear model is constructed using the electrochemical model SP+ based on the lithium-ion battery, the state of charge (SOC) of the lithium battery is used as the system state variable, the current is used as the input, and the terminal voltage of the lithium battery is output;

[0065] The structure of AEKF algorithm is as follows Figure 2The AEKF algorithm runs once per second in the lifetime state-of-charge (SOC) estimation algorithm and continuously adjusts its parameters. After each parameter adjustment, the SOC estimation parameter values ​​are updated to adapt to the changes in lithium-ion battery aging over time.

[0066] Use LSTM neural network to estimate the health status SOH of lithium batteries;

[0067] In specific implementation, as a preferred embodiment of the present invention, the LSTM neural network introduces memory units and a gating mechanism to estimate the health state SOH of the lithium battery;

[0068] The gating mechanism includes input gate, forget gate and output gate. The mathematical expression of forget gate is:

[0069]

[0070] Among them, f t represents the forget gate, σ represents the Sigmoid activation function, W f represents the forget gate weight matrix, h t-1 Represents the hidden state of the previous moment, x t represents the input at time t, b f Represents the bias vector matrix.

[0071] The mathematical expression of the input gate is:

[0072]

[0073] Among them, i t represents the input gate, represents the candidate memory cell state, W C represents the weight matrix, b i 、b C Represents the bias vector matrix.

[0074] The mathematical expression of the memory unit is:

[0075]

[0076] Among them, C t Represents the memory cell state of LSTM at time t.

[0077] The mathematical expression of the output gate is:

[0078] o t =σ(W o ·[h t-1 ,x t ]+b o )h t =o t tanh(Ct )

[0079] Among them, t represents the output gate, W o represents the output gate weight matrix, b o represents the bias vector matrix, h t represents the hidden state at time t.

[0080] The calculation process of SOH estimation based on LSTM neural network is as follows Figure 3 As shown. Factors such as the battery's lower cutoff voltage during constant-current and constant-voltage charging, the voltage difference between 20% and 80% state of charge, the average rate of change, and the integrated area of ​​the voltage curve can all reflect the degree of battery aging to a certain extent. Therefore, the present invention uses these four eigenvalues ​​as input to the LSTM neural network to ultimately obtain the battery's SOH.

[0081] The input of the LSTM neural network is updated by the estimated state of charge (SOC), and the parameters of the Kalman filter (AEKF) are corrected by the estimated state of health (SOH) to build a joint estimation model. Based on the existing AEKF (Adaptive Extended Kalman Filter) estimator and LSTM (Long Short-Term Memory Network) estimator, the input of the LSTM estimator is updated by the SOC estimated by AEKF, and the model parameters used by the AEKF estimator are corrected by the SOH estimated by LSTM. The specific joint estimation algorithm framework is as follows: Figure 4 shown.

[0082] In specific implementation, as a preferred embodiment of the present invention, the joint estimation model specifically includes:

[0083] The input data includes the charge and discharge time series data of the lithium battery X = (x1, x2, ... x t ); after normalization, the eigenvalues ​​are scaled to the same scale and divided into time windows of fixed length; the LSTM algorithm is used to capture the long-term dependencies in the lithium battery charge and discharge data. The hidden state update formula of the LSTM algorithm is:

[0084] h t =LSTM(x t ,h t-1 )

[0085] Among them, x t is the input at time t, h t-1 It is the hidden state of the previous moment, adopts a multi-layer LSTM structure, and adds a Dropout layer to prevent overfitting.

[0086] The output of the LSTM algorithm is passed through a shared fully connected layer to extract features:

[0087] z=ReLU(W h ·h T +b h )

[0088] Among them, z represents the output feature vector, h T represents the hidden state of LSTM, W h represents the weight of the shared layer, b h Represents the bias of the shared layer; the output of the shared layer is mapped to the estimated values ​​of SOC and SOH through two independent fully connected layers:

[0089] SOC=σ(W soc z+b soc )

[0090] SOH=σ(W soh z+b soh )

[0091] Among them, W soc represents the weight of the SOC branch, b soc represents the bias of the SOC branch; W soh represents the weight of the SOH branch, b soh represents the bias of the SOH branch, and σ is the Sigmoid activation function.

[0092] The loss function of the joint estimation model is calculated and the joint estimation model is trained to achieve joint estimation of the state of charge (SOC) and the state of health (SOH).

[0093] In specific implementation, as a preferred embodiment of the present invention, the loss function of the joint estimation model is calculated:

[0094] The loss function L of the joint estimation model is the weighted sum of SOC loss and SOH loss:

[0095] L=α·L soc +(1-α)·L soh

[0096] Among them, L soc represents the mean square error of SOC, L soh represents the mean square error of SOH, α is the weight parameter;

[0097] The Adam optimizer is used for model training, and the initial learning rate is 10 -3， And use early stopping strategy to prevent overfitting.

[0098] Dropout layers were added between LSTM layers, and L2 regularization was added to the loss function to improve the generalization ability of the model. After determining the joint estimation algorithm, the experiment used 70% of the lithium-ion battery aging data as the training set, 10% as the validation set, and 20% as the test set.

[0099] Example

[0100] like Figure 2 As shown, the present invention adopts the SP+ model combined with the AEKF algorithm when estimating SOC. The SP+ model is an advanced battery model that can perform voltage simulation based on the input current value, the current SOC state, and a series of model parameters. The AEKF (Adaptive Extended Kalman Filter) module is an adaptive extended Kalman filter specifically used to accurately estimate the SOC of lithium-ion batteries. Through the collaborative work of these two modules, the algorithm can achieve high-precision SOC estimation throughout the life cycle of the lithium-ion battery, thereby providing reliable data support for the battery management system. In actual operation, the SOC estimation AEKF will run once per second throughout the entire life SOC estimation algorithm and continuously adjust the parameters. After each parameter adjustment, the SOC estimation AEKF will update its parameter values ​​to adapt to changes in lithium-ion batteries as they age over time.

[0101] The calculation process of SOH estimation based on LSTM neural network is as follows Figure 3 The present invention uses the battery's lower cut-off voltage during constant current and constant voltage charging, the voltage difference corresponding to the 20%-80% state of charge process, the average rate of change, and the integral area of ​​the voltage curve as inputs to the LSTM neural network, and ultimately obtains the battery's SOH.

[0102] like Figure 4 As shown in the figure, based on the existing AEKF (Adaptive Extended Kalman Filter) estimator and LSTM (Long Short-Term Memory) estimator, the two estimators are linked: the AEKF-estimated SOC is used to update the LSTM estimator input, and the LSTM-estimated SOH is used to correct the model parameters used by the AEKF estimator. This mutual updating method can effectively correct model parameter errors caused by battery aging, thereby improving the estimation accuracy of the model-based AEKF algorithm throughout the battery life.

[0103] like Figure 5Figure 2 shows the battery terminal voltage test results at different cycles, performed on batteries at room temperature after 40 and 80 cycles of NEDC testing. As can be seen from the figure, as the battery ages, the terminal voltage decreases under the same current and temperature conditions, and the corresponding discharge end time is shortened. This phenomenon is directly caused by the continuous decline in battery capacity. Therefore, the impact of battery aging must be considered when estimating SOC.

[0104] The SOC estimation accuracy of the SOC estimation algorithm based on AEKF, the SOC estimation algorithm with online update of model parameters, and the SOC and SOH joint estimation algorithm with online update of model parameters are compared under different cycles. Figure 6 shown.

[0105] The root mean square error statistics of SOC estimation using different methods are shown in Table 1.

[0106] Table 1 SOC estimation errors of different methods considering aging state

[0107] The number of cycles is 40 The number of cycles is 80 No parameter update 3.6％ 3.9％ Parameter online update 1.2％ 2.9％ Parameters updated and jointly estimated 0.76％ 0.74％

[0108] The comparison between the SOH value estimated by the LSTM model and the measured value is as follows: Figure 7 According to the error data statistics, the maximum error between the SOH estimated value and the measured value is 6.29%, and the average error is 1.6%.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lithium battery SOC / SOH joint estimation method based on a parameter update mechanism model and an LSTM algorithm, characterized in that: include: Use Kalman filter AEKF algorithm to estimate the state of charge SOC of lithium battery; Use LSTM neural network to estimate the health status SOH of lithium batteries; The input of the LSTM neural network is updated by the estimated state of charge (SOC), and the parameters of the Kalman filter (AEKF) are corrected by the estimated state of health (SOH) to build a joint estimation model. The loss function of the joint estimation model is calculated and the joint estimation model is trained to achieve joint estimation of the state of charge (SOC) and the state of health (SOH).

2. The lithium battery SOC / SOH joint estimation method based on the parameter update mechanism model and LSTM algorithm according to claim 1 is characterized in that: When estimating the state of charge (SOC), a nonlinear model is constructed using the electrochemical model SP+ based on the lithium-ion battery, the state of charge (SOC) of the lithium battery is used as the system state variable, the current is used as the input, and the terminal voltage of the lithium battery is output; The Kalman filter AEKF algorithm is run once per second in the full life state of charge (SOC) estimation algorithm and continuously adjusts parameters; After each parameter adjustment, the parameter value of the state of charge (SOC) estimation is updated to adapt to the changes in the lithium-ion battery over time.

3. The lithium battery SOC / SOH joint estimation method based on the parameter update mechanism model and LSTM algorithm according to claim 1 is characterized in that: The LSTM neural network introduces memory units and gating mechanisms to estimate the health status (SOH) of lithium batteries; The gating mechanism includes an input gate, a forget gate, and an output gate. The mathematical expression of the forget gate is: Among them, f t represents the forget gate, σ represents the Sigmoid activation function, W f represents the forget gate weight matrix, h t-1 represents the hidden state at the previous moment, x t represents the input at time t, b f represents the bias vector matrix; The mathematical expression of the input gate is: i t =σ(W i ·[h t-1 ,x t ]+b i ) Among them, i t represents the input gate, represents the candidate memory cell state, W C represents the weight matrix, b i 、b C represents the bias vector matrix; The mathematical expression of the memory unit is: Among them, C t Represents the memory cell state of LSTM at time t; The mathematical expression of the output gate is: o t =σ(W o ·[h t-1 ,x t ]+b o )h t =o t ·tanh(C t ) Among them, t represents the output gate, W o represents the output gate weight matrix, b o represents the bias vector matrix, h t represents the hidden state at time t.

4. The lithium battery SOC / SOH joint estimation method based on the parameter update mechanism model and LSTM algorithm according to claim 1 is characterized in that: The joint estimation model specifically includes: The input data includes the charge and discharge time series data of the lithium battery X = (x1, x2, ... x t ); after normalization, the eigenvalues ​​are scaled to the same scale and divided into time windows of fixed length; the LSTM algorithm is used to capture the long-term dependencies in the lithium battery charge and discharge data. The hidden state update formula of the LSTM algorithm is: h t =LSTM(x t ,h t-1 ) Among them, x t is the input at time t, h t-1 It is the hidden state of the previous moment, adopts a multi-layer LSTM structure, and adds a Dropout layer to prevent overfitting. The output of the LSTM algorithm is passed through a shared fully connected layer to extract features: z=ReLU(W h ·h T +b h ) Among them, z represents the output feature vector, h T represents the hidden state of LSTM, W h represents the weight of the shared layer, b h Represents the bias of the shared layer; the output of the shared layer is mapped to the estimated values ​​of SOC and SOH through two independent fully connected layers: SOC=σ(W soc ·z+b soc ) SOH=σ(W soh ·z+b soh ) Among them, W soc represents the weight of the SOC branch, b soc represents the bias of the SOC branch; W soh represents the weight of the SOH branch, b soh represents the bias of the SOH branch, and σ is the Sigmoid activation function.

5. The lithium battery SOC / SOH joint estimation method based on the parameter update mechanism model and LSTM algorithm according to claim 1 is characterized in that: Calculate the loss function of the joint estimation model: The loss function L of the joint estimation model is the weighted sum of SOC loss and SOH loss: L=α·L soc +(1-a)·L soh Among them, L soc represents the mean square error of SOC, L soh represents the mean square error of SOH, α is the weight parameter; The Adam optimizer is used for model training, and the early stopping strategy is used to prevent overfitting.

Images