SOC-SOH joint estimation method and system based on biLSTM

Abstract

The invention relates to a biLSTM-based SOC-SOH joint estimation method and system. The method comprises the following steps of obtaining voltage, temperature, current, SOC and SOH time sequence data of an energy storage device in a current cycle period; inputting the voltage and temperature time sequence data into a temperature and voltage prediction sub-network, and outputting a voltage prediction curve and a temperature prediction curve of a next cycle period; inputting the voltage prediction curve, the temperature prediction curve and the current time sequence data into an SOC prediction sub-network, and outputting an SOH prediction curve; and inputting the voltage prediction curve, the temperature prediction curve, the current time sequence data and the SOH prediction curve into an SOH prediction sub-network, and outputting an SOC prediction curve. Compared with the prior art, the method has the advantages of high estimation precision and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of SOC-SOH estimation, and in particular to a biLSTM-based SOC-SOH joint estimation method and system. Background Art

[0002] Lithium batteries have the characteristics of high energy density, long cycle life and low self-discharge rate. They have been widely used as a common energy storage form in energy storage power stations in power grids. However, as lithium batteries age, their internal resistance increases, the risks of thermal runaway and explosion increase, and the energy management system can ensure that the battery is within a safe operating range through the state of charge (SOC) and the state of health (SOH). At present, the research on online state estimation of SOC and SOH is mainly divided into three categories: based on chemical models, based on equivalent circuits, and based on historical data. The first two require professional knowledge to assist in building models, which is difficult and cumbersome. The last one mostly uses deep learning technology for estimation, but most of them estimate SOC and SOH separately, and do not consider the correlation between SOC and SOH. Therefore, the estimation accuracy of SOC and SOH prediction values ​​is affected to a certain extent. Patent application CN115291113A discloses a unified joint estimation method and system for SOC, SOH and RUL of lithium-ion batteries. In this method, the SOC, SOH and RUL prediction models of lithium-ion batteries are used for estimation. The prediction models of SOC, SOH and RUL of lithium-ion batteries include SOH-RUL prediction model and SOC prediction model. During the model training process, the intermediate features of SOH-RUL prediction model are extracted as battery aging information to calibrate the battery dynamic state and capacity of SOC prediction model. Although this method takes into account the correlation between SOC and SOH, the prediction of SOC and SOH depends largely on the understanding of battery aging. The model needs to accurately capture these mechanisms, otherwise the prediction results will be affected.

[0003] Therefore, technicians in this field are committed to developing a stable and reliable method to improve the level of energy storage state estimation, so as to provide a reliable basis for the next step of inspection and operation and maintenance. Summary of the invention

[0004] The purpose of the present invention is to provide a biLSTM-based SOC-SOH joint estimation method and system for improving the prediction accuracy of SOH and SOC.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A biLSTM-based SOC-SOH joint estimation method comprises the following steps:

[0007] The voltage, temperature, current, SOC and SOH time series data of the energy storage device in the current cycle are obtained, and input into the pre-trained biLSTM-based energy storage state prediction model, and the SOH prediction curve and the SOC prediction curve are output, wherein the biLSTM-based energy storage state prediction model includes three sub-networks, namely, a temperature and voltage prediction sub-network, a SOH prediction sub-network and a SOC prediction sub-network;

[0008] The execution process of the biLSTM-based energy storage state prediction model includes:

[0009] Inputting voltage and temperature time series data into the temperature and voltage prediction subnetwork, and outputting a voltage prediction curve and a temperature prediction curve for the next cycle;

[0010] Input the voltage prediction curve, temperature prediction curve and current time series data into the SOC prediction subnetwork, and output a SOH prediction curve;

[0011] The voltage prediction curve, temperature prediction curve, current time series data and SOH prediction curve are input into the SOH prediction subnetwork, and the SOC prediction curve is output.

[0012] Furthermore, the SOC and SOH satisfy the following relationship:

[0013]

[0014] In the formula, Q is the current power of the energy storage device, and C0 is the nominal capacity of the energy storage device.

[0015] Furthermore, it also includes a preprocessing step, wherein the preprocessing step is to normalize the voltage, temperature, current, SOC and SOH time series data.

[0016] Further, the normalization method includes one of minimum-maximum normalization, Z score normalization, maximum normalization and L1 / L2 norm normalization, wherein the calculation expression of the minimum-maximum normalization is:

[0017]

[0018] Where, X norm is the normalized data, X is the original data, and X min , X max are the minimum and maximum values ​​in the data respectively.

[0019] Furthermore, the biLSTM-based energy storage state prediction model also includes a shared information layer, which is connected to the temperature and voltage prediction subnetwork, the SOC prediction subnetwork and the SOH prediction subnetwork respectively. The shared information layer includes multiple bidirectional LSTM units. The shared information layer is used to capture the fluctuation relationship between the input time series data and assist each subnetwork in making predictions. The fluctuation relationship between the input time series data includes the fluctuation relationship between voltage, temperature and current, SOC, SOH, the fluctuation relationship between SOH and voltage, temperature and current time series data, and the fluctuation relationship between SOC and voltage, temperature, SOH and current.

[0020] Furthermore, during the execution of the biLSTM-based energy storage state prediction model, the input of the temperature and voltage prediction subnetwork also includes the fluctuation relationship between the voltage, temperature and current, SOC, and SOH; the input of the SOH prediction subnetwork also includes the fluctuation relationship between SOH and the voltage, temperature, and current time series data; the input of the SOC prediction subnetwork also includes the fluctuation relationship between SOC and the voltage, temperature, SOH, and current.

[0021] Furthermore, the training process of the biLSTM-based energy storage state prediction model is specifically as follows:

[0022] Construct a storage state prediction model based on biLSTM;

[0023] Set the number of iterations and learning rate;

[0024] Set the gradient threshold and regularization factor;

[0025] Acquire historical samples of the energy storage device, wherein the historical samples include voltage, current, temperature, SOC, and SOH time series data;

[0026] The historical samples are normalized and preprocessed as feature quantities and input into the biLSTM model for training to obtain a trained biLSTM-based energy storage state prediction model.

[0027] Furthermore, during the training process, the temperature voltage prediction subnetwork, the SOC prediction subnetwork and the SOH prediction subnetwork are all trained using the same loss function, and the loss function includes one of a mean square error loss function, a root mean square error loss function, a mean absolute error loss function and a Huber loss function, wherein the expression of the mean square error loss function is:

[0028]

[0029] In the formula, MSE is the mean square error, n is the number of samples, and y i is the actual observed value of the ith sample, is the model's predicted value for the i-th sample.

[0030] Furthermore, when predicting the SOC, the SOH prediction curve as input data is constant in a single charge and discharge cycle.

[0031] The present invention also provides a biLSTM-based energy storage SOC-SOH joint estimation system, comprising:

[0032] Acquisition module: used to obtain the voltage, temperature, current, SOC and SOH timing data of the energy storage device in the current cycle;

[0033] Voltage and temperature prediction module: used to input voltage and temperature time series data into the temperature and voltage prediction subnetwork, and output the voltage prediction curve and temperature prediction curve of the next cycle;

[0034] SOH prediction module: used to input the voltage prediction curve, temperature prediction curve and current time series data into the SOC prediction subnetwork, and output the SOH prediction curve;

[0035] SOC prediction module: used to input the voltage prediction curve, temperature prediction curve, current time series data and SOH prediction curve into the SOH prediction subnetwork, and output the SOC prediction curve.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] (1) Considering that there is a correlation between the state of health SOH and the state of charge SOC of the energy storage, and the values ​​of SOH and SOC are related to time, temperature, current and voltage, the present invention adopts a biLSTM-based energy storage state prediction model to predict data with time series characteristics. First, the temperature and voltage of the next cycle are predicted, and then the SOH is predicted. Finally, the SOC is predicted through the predicted curve, which can effectively improve the prediction accuracy of SOH and SOC.

[0038] (2) The present invention takes into account that the energy storage health state SOH is one of the main factors affecting the energy storage SOC. The SOH changes at a slow rate and has a theoretical connection with the SOC. Therefore, when predicting the SOC, the SOH is considered to be unchanged. The SOH prediction curve is used as the input when predicting the SOC, which can more accurately predict the SOC curve.

[0039] (3) The biLSTM-based energy storage state prediction model of the present invention also includes a shared information layer for capturing the fluctuation relationship between the input sequence data, so as to better learn the intrinsic connection between SOC and SOH, especially when the training data is limited, which can effectively reduce the risk of overfitting of the model.

[0040] (4) The present invention does not require a mechanism to capture aspects such as battery aging, has fast calculation speed and high accuracy, and is suitable for practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0042] Figure 2 It is a diagram of the internal structure of the energy storage device of the present invention;

[0043] Figure 3 is an equivalent circuit diagram of a lithium-ion battery of the present invention;

[0044] Figure 4 This is the biLSTM principle diagram of the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0046] Example 1

[0047] This embodiment provides a SOC-SOH joint estimation method based on biLSTM. Figure 1 As shown, the method comprises the following steps:

[0048] S1. Analyze the correlation between SOC and SOH and normalize the data.

[0049] The energy storage device used in this embodiment is as follows Figure 2 As shown, the energy storage device is provided with multiple energy storage units, each of which contains a lithium-ion battery, which supplies power to the load grid through DC / DC (bidirectional direct current converter) and DC / AC (direct current alternating current converter). The equivalent circuit of the lithium-ion battery is as follows: Figure 3 As shown in the figure, R0, R1, R2 represent resistors, C1, C2 represent capacitors, OCV, U P , U represent voltage drop, polarization voltage drop, and output voltage respectively.

[0050] Obtain historical samples of energy storage devices, including voltage, current, temperature, SOC, SOH time series data and other characteristic quantities. The correlation formula between SOC and SOH of energy storage devices is:

[0051]

[0052] Where Q is the current charge of the lithium battery energy storage device; C0 is the nominal capacity of the energy storage device.

[0053] SOH changes during the charging and discharging process of energy storage, so SOH has a non-negligible impact on SOC estimation, so it is necessary to adopt a SOC-SOH joint estimation method.

[0054] Normalize the samples to eliminate the impact of different orders of magnitude and units between different features.

[0055] The normalization method may be one of minimum-maximum normalization, Z-score normalization, maximum normalization, and L1 / L2 norm normalization. In this embodiment, the minimum-maximum normalization method is used, and the calculation expression is:

[0056]

[0057] Where, X norm is the normalized data, X is the original data, and X min , X max are the minimum and maximum values ​​in the data respectively.

[0058] S2. Construct a storage state prediction model based on the biLSTM weight sharing multi-task learning framework.

[0059] In battery management systems, the changes in SOC and SOH are usually affected by complex time series, including historical usage data, environmental conditions, etc. In order to improve the accuracy of energy storage state prediction, this implementation introduces a multi-task learning framework with biLSTM weight sharing to better utilize the correlation between them in the joint prediction of SOC and SOH. By sharing some network parameters, the framework enables the model to use training data more efficiently when predicting SOC and SOH, thereby significantly improving the accuracy and robustness of the prediction. Specifically, a shared layer is constructed on the basis of the traditional bidirectional long short-term memory (biLSTM) network. This layer consists of multiple bidirectional LSTM units, which are responsible for capturing the common features in the input sequence, including the fluctuation relationship between voltage, temperature and current, SOC, and SOH, the fluctuation relationship between SOH and voltage, temperature, and current time series data, and the fluctuation relationship between SOC and voltage, temperature, SOC, and current time series data. The parameters of the shared layer are information-shared in the SOC and SOH prediction tasks. This design helps the model share information between two related tasks, so as to better learn the intrinsic connection between SOC and SOH. Especially when the training data is limited, the use of shared features can effectively reduce the risk of overfitting of the model. After the shared layer, the model also includes three sub-networks connected to it respectively: temperature and voltage prediction sub-network, SOC prediction sub-network and SOH prediction sub-network. These three sub-networks contain dedicated bidirectional LSTM layers respectively to further extract specific features of their respective tasks. In this way, the model can not only utilize the common features of the shared layer, but also perform specialized feature extraction for each task in the task-specific layer. This design ensures the independence of SOC and SOH predictions while retaining the synergy between them, so that the overall prediction ability of the model is optimized.

[0060] biLSTM is a neural network model for modeling sequence data, especially for fields such as natural language processing. It combines the characteristics of long short-term memory (LSTM) and bidirectional recurrent network (bidirectional RNN). In biLSTM, the input sequence is first processed by two directional LSTM layers: one from left to right and one from right to left. The two directional LSTM layers learn the feature representation of each position in the input sequence respectively. For each time step, the two directional LSTM layers output a hidden state, which represents the context information before and after the current time step respectively. These hidden states are then concatenated or merged to form a more comprehensive representation. Finally, the output of biLSTM can be passed to other layers, such as fully connected layers for classification or regression, according to the task requirements. Using biLSTM networks can capture both forward and backward information in sequence data, which is particularly important for improving the prediction accuracy of SOC and SOH.

[0061] like Figure 4 As shown in the figure, biLSTM can process both forward and reverse time series at the same time because it has two LSTM structures. The formula is as follows:

[0062]

[0063] where x t is the input information at time t; are the weights of the forward and reverse LSTM respectively; b y is bias; They are the hidden states of the forward LSTM output at time t-1 and t respectively. They are the hidden states of the reverse LSTM output at time t-1 and time t, y t is the output value at time t.

[0064] biLSTM can capture features in different directions in the sequence, which helps to improve the model's ability to extract complex patterns and features in sequence data, thereby improving the model's prediction performance. The temperature and voltage prediction subnetwork, SOC prediction subnetwork, and SOH prediction subnetwork in the biLSTM-based energy storage state prediction model built in this implementation use the same loss objective function for training. The three subnetworks can be trained using the mean square error (MSE) loss function, the root mean square error (RMSE) loss function, the mean absolute error (MABS) loss function, or the Huber loss function. The mean square error (MSE) loss function is used in this example, and the expression is:

[0065]

[0066] Where n is the number of samples, y iis the actual observed value of the ith sample, is the model's predicted value for the i-th sample.

[0067] The historical samples are divided into training set, validation set and test set. The training set is used for training, and the number of iterations, learning rate, gradient threshold and regularization factor of the network are set until the training is completed to obtain the trained biLSTM-based energy storage state prediction model. After the training is completed, the validation set and test set are used to verify and test the biLSTM-based energy storage state prediction model.

[0068] S3. Predict the energy storage voltage and temperature.

[0069] When predicting the energy storage voltage and temperature, the historical fluctuation relationship between voltage, temperature and current, SOC, and SOH is extracted from the shared information layer. For example, there is a nonlinear relationship between voltage and SOC. During the charging process, the voltage will increase with the increase of SOC, and the growth rate will gradually decrease. The above correlation law and the historical fluctuation state of voltage and temperature are used as the input of the energy storage state prediction model to output the energy storage voltage prediction curve and temperature prediction curve in the next cycle.

[0070] S4. Predict SOH based on the voltage prediction curve, the temperature prediction curve and the current timing data, and predict SOC using the voltage, current and SOH as inputs.

[0071] Specifically:

[0072] The fluctuation relationship between the historical SOH and the voltage, temperature, and current time series data is extracted from the shared information layer. For example, there is a negative feedback relationship between current and SOH: high current accelerates SOH decay, and SOH decline affects current stability and battery efficiency. The correlation law, voltage prediction curve, temperature prediction curve, and current value extracted from the shared information layer are input into the SOC prediction subnetwork, and the energy storage SOH curve in the prediction cycle is output.

[0073] Next, the fluctuation relationship between the historical SOC and the voltage, temperature, SOC, and current time series data is extracted from the shared information layer. For example, SOC will drop faster at low temperatures, while SOC will rise faster at high temperatures. The correlation law, voltage prediction curve, temperature prediction curve, current, and SOH prediction curve extracted from the shared information layer are input into the SOH prediction subnetwork, and the energy storage SOC curve in the prediction cycle is output. Because SOH changes more slowly than SOC, it can be considered that when predicting SOC, SOH as input data is unchanged in a single charge and discharge cycle.

[0074] Example 2

[0075] This embodiment provides a biLSTM-based energy storage SOC-SOH joint estimation system, including:

[0076] Acquisition module: used to obtain the voltage, temperature, current, SOC and SOH timing data of the energy storage device in the current cycle;

[0077] Voltage and temperature prediction module: used to input voltage and temperature time series data into the temperature and voltage prediction subnetwork, and output the voltage prediction curve and temperature prediction curve of the next cycle;

[0078] SOH prediction module: used to input the voltage prediction curve, temperature prediction curve and current time series data into the SOC prediction subnetwork, and output the SOH prediction curve;

[0079] SOC prediction module: used to input the voltage prediction curve, temperature prediction curve, current time series data and SOH prediction curve into the SOH prediction subnetwork, and output the SOC prediction curve.

[0080] The rest is the same as in Example 1.

[0081] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0082] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The schemes in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0087] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A SOC-SOH joint estimation method based on biLSTM, characterized in that: The following steps are involved: The voltage, temperature, current, SOC and SOH time series data of the energy storage device in the current cycle are obtained, and input into the pre-trained biLSTM-based energy storage state prediction model, and the SOH prediction curve and the SOC prediction curve are output, wherein the biLSTM-based energy storage state prediction model includes three sub-networks, namely, a temperature and voltage prediction sub-network, a SOH prediction sub-network and a SOC prediction sub-network; The execution process of the biLSTM-based energy storage state prediction model includes: Inputting voltage and temperature time series data into the temperature and voltage prediction subnetwork, and outputting a voltage prediction curve and a temperature prediction curve for the next cycle; Input the voltage prediction curve, temperature prediction curve and current time series data into the SOC prediction subnetwork, and output a SOH prediction curve; The voltage prediction curve, temperature prediction curve, current time series data and SOH prediction curve are input into the SOH prediction subnetwork, and the SOC prediction curve is output.

2. A biLSTM-based SOC-SOH joint estimation method according to claim 1, characterized in that: The SOC and SOH satisfy the following relationship: In the formula, Q is the current power of the energy storage device, and C0 is the nominal capacity of the energy storage device.

3. The SOC-SOH joint estimation method based on biLSTM according to claim 1, characterized in that: The method also includes a preprocessing step, wherein the preprocessing step is to normalize the voltage, temperature, current, SOC and SOH time series data.

4. The SOC-SOH joint estimation method based on biLSTM according to claim 3 is characterized in that: The normalization method includes one of minimum-maximum normalization, Z score normalization, maximum normalization and L1 / L2 norm normalization, wherein the calculation expression of the minimum-maximum normalization is: Where, X norm is the normalized data, X is the original data, and X min , X max are the minimum and maximum values ​​in the data respectively.

5. The SOC-SOH joint estimation method based on biLSTM according to claim 1, characterized in that: The biLSTM-based energy storage state prediction model also includes a shared information layer, which is connected to the temperature and voltage prediction subnetwork, the SOC prediction subnetwork and the SOH prediction subnetwork respectively. The shared information layer includes multiple bidirectional LSTM units. The shared information layer is used to capture the fluctuation relationship between the input time series data and assist each subnetwork in making predictions. The fluctuation relationship between the input time series data includes the fluctuation relationship between voltage, temperature and current, SOC, SOH, the fluctuation relationship between SOH and voltage, temperature and current time series data, and the fluctuation relationship between SOC and voltage, temperature, SOH and current.

6. A biLSTM-based SOC-SOH joint estimation method according to claim 5, characterized in that: During the execution of the biLSTM-based energy storage state prediction model, the input of the temperature and voltage prediction subnetwork also includes the fluctuation relationship between the voltage, temperature and current, SOC, and SOH; the input of the SOH prediction subnetwork also includes the fluctuation relationship between SOH and the voltage, temperature, and current time series data; the input of the SOC prediction subnetwork also includes the fluctuation relationship between SOC and the voltage, temperature, SOH, and current.

7. The SOC-SOH joint estimation method based on biLSTM according to claim 1, characterized in that: The training process of the biLSTM-based energy storage state prediction model is specifically as follows: Construct a storage state prediction model based on biLSTM; Set the number of iterations and learning rate; Set the gradient threshold and regularization factor; Acquire historical samples of the energy storage device, wherein the historical samples include voltage, current, temperature, SOC, and SOH time series data; The historical samples are normalized and preprocessed as feature quantities and input into the biLSTM model for training to obtain a trained biLSTM-based energy storage state prediction model.

8. The SOC-SOH joint estimation method based on biLSTM according to claim 1, characterized in that: During the training process, the temperature and voltage prediction subnetwork, the SOC prediction subnetwork, and the SOH prediction subnetwork are all trained using the same loss function, and the loss function includes one of a mean square error loss function, a root mean square error loss function, a mean absolute error loss function, and a Huber loss function, wherein the expression of the mean square error loss function is: In the formula, MSE is the mean square error, n is the number of samples, and y i is the actual observed value of the ith sample, is the model's predicted value for the i-th sample.

9. The SOC-SOH joint estimation method based on biLSTM according to claim 1, characterized in that: When predicting SOC, the SOH prediction curve used as input data is constant during a single charge and discharge cycle.

10. A biLSTM-based energy storage SOC-SOH joint estimation system, characterized in that: include: Acquisition module: used to obtain the voltage, temperature, current, SOC and SOH timing data of the energy storage device in the current cycle; Voltage and temperature prediction module: used to input voltage and temperature time series data into the temperature and voltage prediction subnetwork, and output the voltage prediction curve and temperature prediction curve of the next cycle; SOH prediction module: used to input the voltage prediction curve, temperature prediction curve and current time series data into the SOC prediction subnetwork, and output the SOH prediction curve; SOC prediction module: used to input the voltage prediction curve, temperature prediction curve, current time series data and SOH prediction curve into the SOH prediction subnetwork, and output the SOC prediction curve.

Images