Training method and device, prediction method, medium and electronic device

Abstract

This application provides a training method and apparatus for a lithium battery state of charge (SOC) prediction model, a lithium battery SOC prediction method, a medium, and an electronic device. The training method includes: constructing a lithium battery SOC prediction model, wherein the lithium battery SOC prediction model is a bidirectional long short-term memory neural network model, whose input parameters include time series of the lithium battery's voltage and current, and whose output parameters include a time series of the lithium battery's SOC; acquiring training data, wherein the training data includes the time series of the lithium battery's voltage, current, and SOC; and training the lithium battery SOC prediction model using the training data. After training, the lithium battery SOC prediction model can accurately predict the lithium battery's SOC.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and relates to a training method, particularly to a training method and apparatus for a lithium battery state of charge prediction model, a lithium battery state of charge prediction method, a medium, and electronic equipment. Background Technology

[0002] With the global energy crisis and the increasing environmental pollution caused by transportation energy consumption, the demand for efficient and clean energy is growing year by year. Lithium-ion batteries, as reusable rechargeable batteries, significantly reduce carbon emissions and chemical fuel consumption. Furthermore, lithium-ion batteries have advantages such as high energy density, low self-discharge rate, and long lifespan, making them widely used in electric vehicles.

[0003] The State of Charge (SOC) of a lithium-ion battery refers to the ratio between the amount of charge currently stored in the battery and its full capacity. SOC can be expressed as a percentage or a decimal, ranging from 0% (fully discharged) to 100% (fully charged). The SOC of a lithium-ion battery is related to its usable energy. By monitoring and understanding the battery's SOC, users can promptly understand the remaining energy and thus rationally plan usage time and charging schedules to avoid the inconvenience of the battery running out at critical moments. Furthermore, SOC is closely related to battery performance and safety. Under different SOC conditions, the battery's output voltage, internal resistance, energy density, and other characteristics will differ. Understanding the performance characteristics of a battery under different SOC conditions allows for a better assessment of its reliability and stability, providing a basis for battery design and application. Moreover, the SOC of a lithium-ion battery is also related to its lifespan. The charge-discharge cycle process of a battery leads to a certain degree of capacity loss and changes in chemical reactions, and these changes are related to the SOC. Properly managing the state of charge (SOC) of a battery and avoiding frequent overcharging or over-discharging can extend its lifespan and reduce the risk of capacity decay and damage. Therefore, accurate monitoring and management of the SOC of lithium batteries is of great importance. Summary of the Invention

[0004] This application provides a training method and apparatus, a prediction method, a medium, and an electronic device for training a lithium battery state of charge prediction model. The trained lithium battery state of charge prediction model can accurately predict the state of charge of the lithium battery.

[0005] In a first aspect, embodiments of this application provide a training method for a lithium battery state of charge prediction model. The training method includes: constructing a lithium battery state of charge prediction model, wherein the lithium battery state of charge prediction model is a bidirectional long short-term memory neural network model, the input parameters of which include the time series of the voltage and current of the lithium battery, and the output parameters of which include the time series of the battery state of charge of the lithium battery; acquiring training data, wherein the training data includes the time series of the voltage, current, and battery state of charge of the lithium battery; and training the lithium battery state of charge prediction model using the training data.

[0006] In one implementation of the first aspect, the lithium battery state of charge prediction model includes an input layer, a bidirectional long short-term memory network layer, a random deactivation layer, a fully connected layer, and an output layer.

[0007] In one implementation of the first aspect, the loss function used in the training is:

[0008]

[0009] Where n is the data length of the time series, and SOC real The SOC is the true value of the state of charge of the lithium battery. simu This is a predicted value for the state of charge of the lithium battery.

[0010] In one implementation of the first aspect, the training method for the lithium battery state of charge prediction model further includes: preprocessing the training data to normalize the training data.

[0011] In one implementation of the first aspect, a Bayesian optimization method is used to optimize the hyperparameters of the lithium battery state of charge prediction model during the training.

[0012] Secondly, embodiments of this application provide a method for predicting the state of charge (SOC) of a lithium battery. The method includes: acquiring time series of voltage and current of the lithium battery; and processing the time series of voltage and current of the lithium battery using a lithium battery SOC prediction model to obtain a time series of the battery SOC of the lithium battery. The lithium battery SOC prediction model is trained using the training method described in any one of the first aspects of this application.

[0013] Thirdly, embodiments of this application provide a training device for a lithium battery state of charge prediction model. The training device includes: a prediction model construction module for constructing a lithium battery state of charge prediction model, wherein the lithium battery state of charge prediction model is a bidirectional long short-term memory neural network model, whose input parameters include time series of the lithium battery voltage and current, and whose output parameters include a time series of the lithium battery state of charge; a training data acquisition module for acquiring training data, wherein the training data includes time series of the lithium battery voltage, current, and battery state of charge; and a prediction model training module for training the lithium battery state of charge prediction model using the training data.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any one of the first aspects of this application.

[0015] Fifthly, embodiments of this application provide an electronic device, the electronic device comprising: a memory storing a computer program; and a processor communicatively connected to the memory, which executes the method described in any one of the first aspects of this application when the computer program is invoked.

[0016] In summary, this application provides a training method for a lithium battery state of charge (SOC) prediction model, used to train the model. After training, the SOC prediction model can be used to accurately predict the SOC of lithium batteries.

[0017] Furthermore, in this embodiment, a bidirectional long short-term memory neural network model is used as the lithium battery state of charge prediction model. This model enhances the comprehensiveness of information by simultaneously acquiring both forward and reverse information of the lithium battery, thus enabling more accurate prediction of the lithium battery state of charge.

[0018] In some embodiments, the Bayesian optimization algorithm is used to match the data features of the lithium battery with the network topology of the lithium battery state of charge prediction model, which is beneficial to further improve the performance of the lithium battery state of charge prediction model. Attached Figure Description

[0019] Figure 1A The diagram shows an application scenario of the training method for the lithium battery state of charge prediction model in this application embodiment.

[0020] Figure 1B The flowchart shown is a training method for a lithium battery state-of-charge prediction model in an embodiment of this application.

[0021] Figure 2A and Figure 2BThe diagram shown is a structural schematic of the lithium battery state of charge prediction model in an embodiment of this application.

[0022] Figure 3 The diagram shown is a schematic representation of the training device 3 for the lithium battery state of charge prediction model in this embodiment of the application.

[0023] Figure 4 The diagram shown is a structural schematic of an electronic device in an embodiment of this application.

[0024] Component designation explanation

[0025] 1. Electronic equipment

[0026] 11 Data Acquisition Module

[0027] 12 processors

[0028] 13 Monitors

[0029] 3 Training device for lithium battery state of charge prediction model

[0030] 31 Prediction Model Building Module

[0031] 32 Training Data Acquisition Module

[0032] 33 Prediction Model Training Module

[0033] 4 Electronic devices

[0034] 41 Memory

[0035] 42 processors

[0036] 43 Monitors

[0037] Steps S11 to S13 Detailed Implementation

[0038] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0039] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the illustrations only show the components related to this application and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0040] The state of charge (SOC) of a lithium-ion battery is one of the most important state variables. It is not only a crucial parameter for measuring the battery's remaining power but also a vital indicator for ensuring the battery's safety and lifespan. Accurately obtaining the SOC of a lithium-ion battery is of great significance. However, the SOC of a lithium-ion battery is difficult to measure directly. Therefore, how to accurately obtain the SOC of a lithium-ion battery has become one of the urgent technical problems that engineers in related fields need to solve.

[0041] To address at least the aforementioned issues, this application provides a training method for a lithium battery state of charge (SOC) prediction model, used to train the model. After training, the SOC prediction model can be used to accurately predict the SOC of lithium batteries. Furthermore, in this application embodiment, a bidirectional long short-term memory (LSTM) neural network model is used as the SOC prediction model. This model enhances the comprehensiveness of the information by simultaneously acquiring both forward and reverse information from the lithium battery, thus enabling more accurate prediction of the SOC.

[0042] The technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0043] Figure 1A This diagram illustrates an application scenario of the training method for the lithium battery state-of-charge prediction model provided in this embodiment of the application. Figure 1A As shown, the electronic device 1 includes a data acquisition module 11, a processor 12, and a display 13. This electronic device 1 can be used to implement the training method for the lithium battery state-of-charge prediction model provided in the embodiments of this application; however, the application scenarios of the training method for the lithium battery state-of-charge prediction model provided in the embodiments of this application are not limited to... Figure 1A Electronic device 1 shown.

[0044] Figure 1B The flowchart shown is a training method for a lithium battery state-of-charge prediction model provided in an embodiment of this application. Figure 1B As shown, the training method for the lithium battery state of charge prediction model includes the following steps S11 to S13.

[0045] S11. Construct a lithium battery state of charge prediction model. The lithium battery state of charge prediction model is a bidirectional long short-term memory neural network model. Its input parameters include the time series of the lithium battery's voltage and current, and its output parameters include the time series of the lithium battery's state of charge.

[0046] S12, acquire training data, which includes the time series of lithium battery voltage, current and battery state of charge.

[0047] Optionally, in step S12, sensors can be used to acquire parameters such as the voltage, current, and state of charge of the lithium battery. Furthermore, in some possible implementations, a precise electrochemical model can also be used to acquire these parameters.

[0048] S13, The lithium battery state of charge (SOC) prediction model is trained using training data. Specifically, during training, the SOC prediction model continuously adjusts its internal parameters and weights to minimize the error between the predicted result and the actual SOC. Through iterative training, the SOC prediction model gradually extracts key features of the battery SOC and establishes a mapping relationship between the time series of the battery's voltage and current and the time series of the battery SOC.

[0049] As described above, the training method provided in this application embodiment can train a lithium battery state of charge (SOC) prediction model. After training, the SOC prediction model can be used to accurately predict the SOC of lithium batteries. Furthermore, in this application embodiment, a bidirectional long short-term memory (LSTM) neural network model is used as the SOC prediction model. This model enhances the comprehensiveness of the information by simultaneously acquiring both forward and reverse information from the lithium battery, thus enabling more accurate prediction of the SOC.

[0050] Please see Figure 2A and Figure 2B In one embodiment of this application, the lithium battery state of charge prediction model includes an input layer, a bidirectional long short-term memory network layer, a random deactivation layer, a fully connected layer, and an output layer.

[0051] The input layer is the starting point of the lithium-ion battery state-of-charge (SOC) prediction model. It receives input data from external sources, including information related to the battery's SOC, such as voltage and current. The input layer then passes this data to the next layer for processing and analysis. In some possible implementations, the input layer is x. t ={V t ,I t}, where V t For the time series of voltage, I t Let t be the time series of current, where t represents time.

[0052] The bidirectional Long Short-Term Memory (LSTM) network layer is a core component of the lithium-ion battery state-of-charge (SOC) prediction model. LSM is a recurrent neural network with memory units and gating mechanisms, enabling efficient processing of sequential data. By simultaneously considering past and future contextual information, the bidirectional LSM network layer can better capture the dynamic changes and dependencies in the battery's SOC. In this embodiment, the bidirectional LSM network layer includes two independent layers, one for the forward sequence and the other for the reverse sequence. This bidirectional structure compensates for the lack of information in the network, thereby better capturing the long-term contextual dependencies in time-series tasks and improving the accuracy of the prediction results. Figure 2B middle, SOC(t) = w0 × h f +b0. Among them, and These represent the values ​​of the hidden layer at different times during forward propagation. and These represent the values ​​of the hidden layer at different times during backpropagation, h. f This represents the output of the fully connected layer, where w0 and b0 represent the weight matrix and bias of the fully connected layer, respectively.

[0053] A random deactivation layer is used in each training batch by setting half of the hidden layer node values ​​to 0, thereby reducing the risk of overfitting by ignoring a certain number of feature detectors. By configuring a random deactivation layer in the lithium-ion battery state-of-charge prediction model, the interaction between feature detectors (hidden layer nodes) can be reduced. This interaction refers to the dependence of some detectors on others to function. This approach helps improve the model's generalization ability and robustness, allowing it to better adapt to unseen data.

[0054] In some possible implementations, a random deactivation layer can be appended after the fully connected layer to randomly deactivate neurons with a 50% probability, thereby improving the accuracy of the predicted battery state of charge.

[0055] It should be noted that in some embodiments, the lithium battery state of charge prediction model may not include a random deactivation layer.

[0056] Fully connected layers are used for feature extraction and transformation. They perform linear combinations and nonlinear transformations on the input features to produce higher-level feature representations. Neurons in fully connected layers are connected to all neurons in the previous layer, and through the adjustment of weights and biases, more complex representations and learning of the input data are achieved.

[0057] The output layer is the final layer of the lithium-ion battery state-of-charge (SOC) prediction model, used to transform the model's output into a predicted SOC result. Depending on the task, the output layer may employ different activation and loss functions to adapt to different prediction requirements.

[0058] In some possible implementations, the connections and parameter adjustments between the aforementioned layers can be optimized using a backpropagation algorithm, enabling the lithium-ion battery state-of-charge (SOC) prediction model to learn and adjust based on the input data. It should be understood that the structure and parameter settings of the lithium-ion battery SOC prediction model can be designed and optimized according to specific tasks and data characteristics to achieve accurate battery SOC prediction.

[0059] In one embodiment of this application, the training method for the lithium battery state of charge prediction model may further include: preprocessing the training data to normalize the training data. Specifically, since the collected training data have different magnitudes and different dimensions, these differences can affect the training speed and training effect of the model. To address this issue, data preprocessing is used in this embodiment to eliminate the adverse effects caused by these differences.

[0060] In some possible implementations, the data samples can be normalized using the following formula:

[0061]

[0062] Where, x * This refers to the normalized data, where x refers to the actual collected data, including lithium battery voltage, current, and state of charge. min and x max This refers to the maximum and minimum values ​​in the actual collected data.

[0063] In some possible implementations, the training method for the lithium-ion battery state-of-charge (SOC) prediction model may also include: dividing the collected voltage, current, and SOC data into a training set and a test set. The training set includes training data used to train the lithium-ion battery SOC prediction model. The test set includes test data used to test the lithium-ion battery SOC prediction model.

[0064] In one embodiment of this application, the loss function used in training the lithium battery state of charge prediction model using training data is:

[0065]

[0066] Where n is the data length of the time series, and SOC real The SOC is the true value of the state of charge of a lithium battery. simuThis is a predicted value for the state of charge of the lithium battery.

[0067] In one embodiment of this application, when testing a lithium battery state-of-charge prediction model using data from a test set, the following formula can be used as an evaluation index:

[0068] In one embodiment of this application, Adam (Adaptive Moment Estimation) can be used to train the lithium battery state of charge prediction model.

[0069] In one embodiment of this application, a Bayesian optimization method is used to optimize the hyperparameters of the lithium battery state of charge (SOC) prediction model during training. Specifically, the generalization ability of the SOC prediction model is related to several hyperparameters, such as the number of hidden layer neurons, the maximum number of iterations, the initial learning rate, and the learning rate decay factor. For example, if the number of hidden layer neurons is too large, overfitting may occur. If the number of hidden layer neurons is too small, it will be difficult to obtain sufficient information during training. To address this issue, this embodiment uses a Bayesian optimization method to optimize the hyperparameters of the SOC prediction model. Bayesian optimization can combine previous evaluation results, achieving better results with fewer iterations. The core of optimizing the hyperparameters of the SOC prediction model using Bayesian optimization is to estimate the posterior distribution of the objective function using Bayes' theorem, and then select the next combination of hyperparameters to be sampled based on this distribution. This method continuously optimizes the objective function using the sampling results of the previous sampling point until the globally optimal hyperparameters are found.

[0070] In some possible implementations, the hyperparameters are combined as S = s1, s2, ..., s n The optimal combination of hyperparameters is:

[0071]

[0072] Among them, s * Let represent the optimal combination of hyperparameters, and f(s) represent the objective function with respect to hyperparameters s.

[0073] The following section will detail the process of optimizing the hyperparameters of the lithium battery state of charge prediction model using the Bayesian optimization method. This process specifically includes the following steps S31 to S34.

[0074] S31 defines the search space and optimization objective for the hyperparameters. The search space is the range of possible values ​​for the hyperparameters.

[0075] S32, Construct a Gaussian process regression model to model the relationship between hyperparameters and model performance. Gaussian process regression fits a Gaussian distribution to the training data as a probabilistic model between hyperparameters and model performance. This model provides a performance prediction and the uncertainty of the prediction at each point in the hyperparameter space.

[0076] In each iteration of the Bayesian optimization, the Gaussian process regression model is first updated using existing training data to obtain the performance predictions and uncertainties at each point in the current hyperparameter space. Then, based on a selection strategy, such as using expected improvement or confidence upper bounds, the next exploration point, i.e., the next candidate hyperparameter, is selected. This exploration point can be determined based on the predictions and uncertainties of the Gaussian process regression model.

[0077] S34. Train the model using the selected candidate hyperparameters to obtain the corresponding model performance. Add this new data point to the training dataset and update the Gaussian process regression model.

[0078] By iterating through the above process, the Bayesian optimization method can find a better combination of hyperparameters within a finite number of iterations, thereby improving the performance of the state of charge prediction model. This approach fully utilizes historical observation data and the uncertainty information of the model, effectively exploring and utilizing the hyperparameter search space, and avoiding overfitting and local optima problems in traditional optimization methods.

[0079] The scope of protection for the lithium battery state of charge prediction method described in this application is not limited to the order of steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application.

[0080] This application also provides a method for predicting the state of charge (SOC) of a lithium battery. The method includes: acquiring time series data of the voltage and current of the lithium battery; and processing the time series data of the voltage and current using a lithium battery SOC prediction model to obtain a time series data of the battery's SOC. The lithium battery SOC prediction model is trained using the training method provided in any embodiment of this application.

[0081] This application also provides a training device for a lithium battery state of charge prediction model. The training device can implement the training method for the lithium battery state of charge prediction model described in this application. However, the implementation device for the training method of the lithium battery state of charge prediction model described in this application includes, but is not limited to, the structure of the training device for the lithium battery state of charge prediction model listed in this embodiment. All structural modifications and substitutions of the prior art made based on the principles of this application are included within the protection scope of this application.

[0082] Figure 3 The diagram shown is a structural schematic of the training device 3 for the lithium battery state-of-charge prediction model provided in an embodiment of this application. Figure 3 As shown, the training device 3 for the lithium battery state of charge (SOC) prediction model includes a prediction model construction module 31, a training data acquisition module 32, and a prediction model training module 33. The prediction model construction module 31 is used to construct the lithium battery SOC prediction model. The lithium battery SOC prediction model is a bidirectional long short-term memory neural network model. Its input parameters include the time series of the lithium battery's voltage and current, and its output parameters include the time series of the lithium battery's SOC. The training data acquisition module 32 is used to acquire training data, which includes the time series of the lithium battery's voltage, current, and SOC. The prediction model training module 33 is used to train the lithium battery SOC prediction model using the training data.

[0083] It should be noted that in the training device 3 for the lithium battery state of charge prediction model provided in this application embodiment, the prediction model construction module 31, the training data acquisition module 32, and the prediction model training module 33 are connected to... Figure 1B Steps S11 to S13 in the training method of the lithium battery state of charge prediction model shown correspond one-to-one, and will not be elaborated on here.

[0084] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.

[0085] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of this application, depending on actual needs. For example, the functional modules / units in the various embodiments of this application may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.

[0086] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0087] This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the training method and / or the lithium battery state-of-charge prediction method of any embodiment of this application. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state hard disk, magnetic tape, floppy disk, optical disk, and any combination thereof. The above storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0088] This application also provides an electronic device. Figure 4 The diagram shown is a structural schematic of electronic device 4 in one embodiment of this application. Figure 4 As shown, in this embodiment, electronic device 4 includes memory 41 and processor 42.

[0089] The memory 41 is used to store computer programs. In some possible implementations, the memory 41 may include various media capable of storing program code, such as ROM, RAM, magnetic disk, USB flash drive, memory card, or optical disk.

[0090] In this embodiment, memory 41 may include a computer system readable medium in the form of volatile memory, such as RAM and / or cache memory. Electronic device 4 may further include other removable / non-removable, volatile / non-volatile computer system storage media. Memory 41 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0091] The processor 42 is connected to the memory 41 and is used to execute the computer program stored in the memory 41 so that the electronic device 4 executes the training method of the lithium battery state of charge prediction model and / or the lithium battery state of charge prediction method.

[0092] In this embodiment, the processor 42 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0093] In some possible implementations, the electronic device 4 provided in this application embodiment may further include a display 43. The display 43 is communicatively connected to the memory 41 and the processor 42, and is used to display the training method of the lithium battery state of charge prediction model and / or the related graphical user interface (GUI) of the lithium battery state of charge prediction method.

[0094] In this embodiment, the display 43 may include a display screen (display panel). In some implementations, the display panel may be configured using a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like. Alternatively, the display 43 may also be a touch panel (touchscreen, touch screen), which may include a display screen and a touch-sensitive surface. When the touch-sensitive surface detects a touch operation on or near it, it transmits the information to the processor 42 to determine the type of touch event. Subsequently, the processor 42 provides corresponding visual output on the display device based on the type of touch event.

[0095] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A training method for a lithium battery state-of-charge prediction model, characterized in that, include: A lithium battery state of charge prediction model is constructed. The lithium battery state of charge prediction model is a bidirectional long short-term memory neural network model. Its input parameters include the time series of the voltage and current of the lithium battery, and its output parameters include the time series of the battery state of charge. The lithium battery state of charge prediction model includes an input layer, a bidirectional long short-term memory network layer, a random deactivation layer, a fully connected layer, and an output layer. The random deactivation layer is used to configure half of the hidden layer node values ​​to 0 in each training batch. Acquire training data, which includes the time series of the lithium battery's voltage, current, and state of charge. The lithium battery state of charge prediction model is trained using the training data; In the training process, Bayesian optimization is used to optimize the hyperparameters of the lithium battery state of charge prediction model, including: Define the search space and optimization objective for the hyperparameters, where the search space is the range of values ​​for the hyperparameters; A Gaussian process regression model is constructed to model the relationship between hyperparameters and model performance; In each iteration of Bayesian optimization, the Gaussian process regression model is updated using the existing training data to obtain the performance prediction and uncertainty of each point in the current hyperparameter space. The next exploration point is selected according to the selection strategy, and the exploration point is selected based on the performance prediction and uncertainty of the Gaussian process regression model. The model is trained using the selected candidate hyperparameters to obtain the corresponding model performance. This new data point is then added to the training dataset, and the Gaussian process regression model is updated.

2. The training method according to claim 1, characterized in that, The lithium battery state of charge prediction model includes an input layer, a bidirectional long short-term memory network layer, a random deactivation layer, a fully connected layer, and an output layer.

3. The training method according to claim 1, characterized in that, The loss function used in the training is: Where n is the data length of the time series, and SOC real The SOC is the true value of the state of charge of the lithium battery. simu This is a predicted value for the state of charge of the lithium battery.

4. The training method according to claim 1, characterized in that, Also includes: The training data is preprocessed to normalize it.

5. A method for predicting the state of charge of a lithium battery, characterized in that, include: Obtain the time series of voltage and current of the lithium battery; The time series of voltage and current of the lithium battery are processed using a lithium battery state of charge prediction model to obtain the time series of the battery state of charge of the lithium battery, wherein the lithium battery state of charge prediction model is trained using the training method described in any one of claims 1 to 4.

6. A training device for a lithium battery state of charge prediction model, characterized in that, include: The prediction model building module is used to build a lithium battery state of charge prediction model. The lithium battery state of charge prediction model is a bidirectional long short-term memory neural network model. Its input parameters include the time series of the voltage and current of the lithium battery, and its output parameters include the time series of the battery state of charge of the lithium battery. The lithium battery state of charge prediction model includes an input layer, a bidirectional long short-term memory network layer, a random deactivation layer, a fully connected layer, and an output layer. The random deactivation layer is used to configure half of the hidden layer node values ​​to 0 in each training batch. The training data acquisition module is used to acquire training data, which includes the time series of the voltage, current and state of charge of the lithium battery. The prediction model training module is used to train the lithium battery state of charge prediction model using the training data; In the training process, Bayesian optimization is used to optimize the hyperparameters of the lithium battery state of charge prediction model, including: Define the search space and optimization objective for the hyperparameters, where the search space is the range of values ​​for the hyperparameters; A Gaussian process regression model is constructed to model the relationship between hyperparameters and model performance; In each iteration of Bayesian optimization, the Gaussian process regression model is updated using the existing training data to obtain the performance prediction and uncertainty of each point in the current hyperparameter space. The next exploration point is selected according to the selection strategy, and the exploration point is selected based on the performance prediction and uncertainty of the Gaussian process regression model. The model is trained using the selected candidate hyperparameters to obtain the corresponding model performance. This new data point is then added to the training dataset, and the Gaussian process regression model is updated.

7. The training device according to claim 6, characterized in that, In the training process, Bayesian optimization is used to optimize the hyperparameters of the lithium battery state of charge prediction model.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

9. An electronic device, characterized in that, The electronic device includes: A memory that stores a computer program; The processor, which is communicatively connected to the memory, executes the method of any one of claims 1 to 6 when the computer program is invoked.

Images