Power battery peak power calibration-free estimation method and device, equipment and medium

Abstract

The application relates to a power battery peak power calibration-free estimation method, device, equipment and medium, which comprises the following steps: fine-tuning the end full connection layer parameters of a first residual compensation model with a low learning rate based on the second low-power operation data of a target domain battery to obtain a second residual compensation model matched with the target domain battery; constructing a second fractional order battery terminal voltage model according to the second low-power operation data of the target domain battery, and correcting the terminal voltage prediction value output by the second fractional order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery; solving a constraint optimization problem to obtain the maximum sustainable constant power of the target domain battery within a preset time window that meets all constraint conditions, as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery. The application can avoid high stress working condition test of the target battery, slow down the battery aging and prolong the service life.

Description

Technical Field

[0001] This application relates to the field of power batteries, and in particular to a method for estimating the peak power of a power battery without calibration, a corresponding device, electronic equipment, and a computer-readable storage medium. Background Technology

[0002] With the widespread application of power batteries in various electric systems, the requirements for accurate estimation of battery state by the Battery Management System (BMS) are constantly increasing. Among them, peak power (SOP) is an important indicator for measuring the maximum sustainable charge and discharge capacity that a battery can provide within a given time window, which directly affects the power supply capacity and safe operating boundary of the system under dynamic load conditions.

[0003] Existing methods for estimating the peak power of power batteries mainly include:

[0004] One approach is the lookup table method based on empirical mapping. This method establishes a mapping relationship between power and variables such as state of charge (SOC) and temperature through a large amount of high-rate or peak power test data. During use, the peak power value under the current operating condition is obtained through interpolation. This method is simple to implement, but it requires calibration experiments on the target battery under multiple rates and temperatures, resulting in high testing costs and long cycles. Furthermore, high-rate boundary testing may adversely affect battery life.

[0005] Secondly, there is the extrapolation method based on physical models. This method typically uses equivalent circuit models or electrochemical models to predict terminal voltage changes under the assumption of a constant current or constant power trajectory, and then inversely deduces the feasible maximum power based on constraints such as the cutoff voltage. However, physical models often suffer from model mismatch problems under high-rate conditions, especially when the battery polarization effect is enhanced or the temperature change is significant, the model error increases, leading to a decrease in the accuracy of peak power estimation.

[0006] Thirdly, data-driven prediction methods utilize nonlinear models such as neural networks to directly establish the mapping relationship between input variables and peak power, exhibiting strong expressive power. However, these methods typically rely on a large amount of high-power or peak-power calibration data for training. Once the battery model or batch is changed, recalibration is required, resulting in limited cross-battery generalization ability.

[0007] In recent years, model-data fusion methods have been proposed to improve the accuracy of peak power estimation. These methods have alleviated the extrapolation error problem of physical models to some extent. However, most still rely on high-rate calibration data of the target battery, making it difficult to avoid repeated high-power testing. In practical engineering applications, different models or batches of batteries often need to undergo repeated high-rate or peak power tests to complete calibration. This process not only increases testing costs and development cycles, but may also accelerate battery aging under high-stress conditions, affecting safety and reliability.

[0008] In summary, adapting to different models or batches of batteries in the existing technology often requires repeated high-rate or peak power tests to complete the calibration. This process not only increases testing costs and development cycles, but may also accelerate battery aging under high-stress conditions, affecting safety and reliability. The applicant has made corresponding explorations to solve this problem. Summary of the Invention

[0009] The purpose of this application is to solve the above-mentioned problems by providing a calibration-free method for estimating the peak power of a power battery, a corresponding device, electronic equipment, and a computer-readable storage medium.

[0010] To achieve the various objectives of this application, the following technical solution is adopted:

[0011] A calibration-free method for estimating the peak power of a power battery, proposed to meet one of the purposes of this application, includes:

[0012] Obtain the first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model.

[0013] Using the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery as sample labels, the first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained to convergence; based on the second low-power operation data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence are fine-tuned with a low learning rate to obtain the second residual compensation model that matches the target domain battery;

[0014] A second fractional-order battery terminal voltage model is constructed based on the second low-power operating data of the target domain battery. The terminal voltage prediction value output by the second fractional-order battery terminal voltage model is corrected by residual compensation based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery.

[0015] Based on the compensated terminal voltage, a constraint optimization problem is constructed, which includes upper and lower limit constraints on the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. The constraint optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery.

[0016] Optionally, the steps for constructing the first fractional-order battery terminal voltage model include:

[0017] Acquire the first low-power operating data of the source domain battery, including battery current, battery state of charge, measured terminal voltage, and battery temperature;

[0018] Based on the battery current, battery state of charge, and battery temperature in the first low-power operation data, a first fractional-order battery terminal voltage model is constructed that uses fractional-order operators to describe the dynamic behavior of the battery terminal voltage. The model parameters of the first fractional-order battery terminal voltage model are determined by minimizing the root mean square error between the predicted terminal voltage output by the first fractional-order battery terminal voltage model and the measured terminal voltage in the first low-power operation data.

[0019] Optionally, the steps for training the first residual compensation model include:

[0020] Obtain a training sample set, wherein the training sample set includes multiple training samples and their corresponding sample labels, the training samples include the first low-power operation data and the first high-power operation data of the source domain battery, and the sample labels characterize the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery.

[0021] Multiple training samples and their corresponding sample labels are input into a residual mapping function constructed by a bidirectional long short-term memory network with a fully connected layer. The network parameters of the residual mapping function are iteratively trained to fit the residual mapping function to fit the variation law of the terminal voltage residual of the source domain battery under low power and high power conditions. This process continues until the first residual compensation model is trained to convergence, thus determining the first residual compensation model that has been trained to convergence.

[0022] Optionally, the step of fine-tuning the parameters of the final fully connected layer of the first residual compensation model, which has been trained to convergence, with a low learning rate based on the second low-power operating data of the target domain battery, to obtain a second residual compensation model matching the target domain battery, includes:

[0023] The second low-power operating data of the target domain battery is input into the first residual compensation model that has been trained to convergence;

[0024] The network parameters of the bidirectional long short-term memory network in the first residual compensation model are fixed, and only the parameters of the terminal fully connected layer of the first residual compensation model are iteratively fine-tuned with a low learning rate. This allows the fine-tuned first residual compensation model to fit the terminal voltage residual variation law under the low power condition of the target domain battery. After the fine-tuning is completed, a second residual compensation model matching the target domain battery is obtained.

[0025] Optionally, the step of performing residual compensation correction on the predicted terminal voltage value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery includes:

[0026] The battery current, battery state of charge, and battery temperature from the second low-power operating data of the target domain battery are input into the second residual compensation model to obtain the terminal voltage residual compensation value of the target domain battery output by the second residual compensation model.

[0027] The terminal voltage prediction value output by the second fractional-order battery terminal voltage model is compensated based on the terminal voltage residual compensation value to obtain the compensated terminal voltage of the target domain battery.

[0028] Optionally, the step of solving for the peak power of the target domain battery includes:

[0029] Based on the compensated terminal voltage, a constrained optimization problem is constructed with the maximum sustainable constant power as the optimization objective, and including upper and lower limits of battery current, terminal voltage, battery state of charge, and battery temperature.

[0030] The constrained optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window. The maximum sustainable constant power is then used as the peak power of the target domain battery.

[0031] Optionally, the first low-power operating data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the source domain battery; the second low-power operating data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the target domain battery.

[0032] The basic network architecture of the first residual compensation model and the second residual compensation model is a bidirectional long short-term memory network connected in series with a fully connected layer.

[0033] A calibration-free estimation device for the peak power of a power battery, provided for another purpose of this application, includes:

[0034] The terminal voltage prediction module is configured to acquire the first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model.

[0035] The compensation model training module is configured to use the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and the terminal voltage residual value between the predicted value and the measured value of the terminal voltage of the source domain battery as the sample label. The first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained to convergence. Based on the second low-power operation data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence are fine-tuned with a low learning rate to obtain the second residual compensation model that matches the target domain battery.

[0036] The terminal voltage compensation module is configured to construct a second fractional-order battery terminal voltage model based on the second low-power operating data of the target domain battery, and perform residual compensation correction on the terminal voltage prediction value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery.

[0037] The peak power estimation module is configured to construct a constraint optimization problem based on the compensated terminal voltage, which includes upper and lower limit constraints on the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. The constraint optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery.

[0038] An electronic device provided for another purpose of this application includes a central processing unit and a memory, wherein the central processing unit is configured to invoke and run a computer program stored in the memory to perform the steps of the peak power estimation method for power batteries described in this application without calibration.

[0039] A computer-readable storage medium is provided for another purpose of this application, which stores, in the form of computer-readable instructions, a computer program implemented according to the power battery peak power calibration-free estimation method, which, when called by a computer, executes the steps included in the corresponding method.

[0040] Compared to existing technologies, this application addresses the issue that existing technologies often require repeated high-rate or peak power tests to calibrate batteries of different models or batches. This process not only increases testing costs and development cycles but may also accelerate battery aging under high-stress conditions, affecting safety and reliability. This application offers the following advantages, including but not limited to:

[0041] Firstly, this application uses transfer learning to transfer the model deviation characteristics of the source domain battery under high power conditions to the target domain battery, enabling the target domain battery to achieve accurate peak power estimation without high power or peak power calibration, reducing the dependence on high rate testing and reducing the testing cost and cycle of power batteries.

[0042] Secondly, this application adopts a combination of fractional-order battery terminal voltage model and residual compensation model, which effectively improves the model mismatch problem at high rates and significantly improves the peak power estimation accuracy of power batteries.

[0043] Thirdly, this application achieves the reuse of source domain battery knowledge to target domain battery through transfer adaptation, improving the generalization ability across battery models and batches; it constructs a multi-constraint safety optimization solution mechanism to ensure that the estimation results are within the safe operating boundary, thereby enhancing the system's reliability and stability; at the same time, it avoids high-stress condition testing of the target power battery, slows down battery aging, and extends its service life, which has extremely strong engineering practical value. Attached Figure Description

[0044] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0045] Figure 1 This is a flowchart illustrating the calibration-free estimation method for the peak power of a power battery in an embodiment of this application.

[0046] Figure 2 This is a flowchart illustrating the overall process of estimating the peak power of the target domain battery in an embodiment of this application.

[0047] Figure 3 This is a schematic diagram of the physical model of the fractional-order battery terminal voltage in the embodiments of this application;

[0048] Figure 4 This is a schematic diagram of the network structure of the residual compensation model in the embodiments of this application;

[0049] Figure 5 This is a comparison chart of the peak power estimation errors of the traditional method and the calibration-free peak power estimation method for power batteries in the embodiments of this application;

[0050] Figure 6 This is a schematic diagram of the peak power estimation device for power batteries without calibration, as described in the embodiments of this application.

[0051] Figure 7 This is a schematic diagram of the structure of the computer device in the embodiments of this application. Detailed Implementation

[0052] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0053] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0054] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0055] Those skilled in the art will understand that the terms "client," "terminal," and "terminal device" as used herein include both devices that receive wireless signals, devices that only possess wireless signal receiver capabilities without transmission capabilities, and devices with receiving and transmitting hardware, devices that have receiving and transmitting hardware capable of bidirectional communication over a bidirectional communication link. Such devices may include: cellular or other communication devices such as personal computers or tablets, having single-line displays, multi-line displays, or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service) that can combine voice, data processing, fax, and / or data communication capabilities; PDAs (Personal Digital Assistants) that may include radio frequency receivers, pagers, internet / intranet access, web browsers, notebooks, calendars, and / or GPS (Global Positioning System) receivers; and conventional laptops and / or handheld computers or other devices that have and / or include radio frequency receivers. As used herein, "client," "terminal," and "terminal device" can be portable, transportable, installed in a means of transportation (air, sea, and / or land), or suitable and / or configured to operate locally and / or in a distributed manner, operating in any other location on Earth and / or in space. "Client," "terminal," and "terminal device" as used herein can also be a communication terminal, an internet access terminal, or a music / video playback terminal, such as a PDA, a MID (Mobile Internet Device), and / or a mobile phone with music / video playback capabilities, or a smart TV, set-top box, etc.

[0056] The hardware referred to by the names "server," "client," and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer. It is a hardware device with the necessary components revealed by the von Neumann architecture, such as a central processing unit (including an arithmetic logic unit and a control unit), memory, input devices, and output devices. The computer program is stored in its memory, and the central processing unit loads the program stored in the secondary storage into the main memory to run it, execute the instructions in the program, and interact with the input and output devices to complete specific functions.

[0057] It should be noted that the concept of "server" used in this application can also be extended to the case of server clusters. Based on the network deployment principles understood by those skilled in the art, the servers should be logically divided. Physically, these servers can be independent of each other but accessible through interfaces, or they can be integrated into a single physical computer or a computer cluster. Those skilled in the art should understand this flexibility and should not use it to constrain the implementation of the network deployment method in this application.

[0058] One or more of the technical features of this application, unless explicitly specified herein, can be deployed on a server and accessed by a client remotely calling the online service interface provided by the server, or can be directly deployed and run on a client for access.

[0059] Unless otherwise specified, the neural network models referenced or potentially referenced in this application may be deployed on a remote server and invoked remotely on the client, or deployed on a client with the capability to invoke directly. In some embodiments, when running on the client, the corresponding intelligence may be acquired through transfer learning in order to reduce the requirements on the client's hardware resources and avoid excessive consumption of the client's hardware resources.

[0060] Unless otherwise specified, all data involved in this application may be stored remotely on a server or on a local terminal device, as long as it is suitable for use by the technical solution of this application.

[0061] Those skilled in the art will understand that although the various methods in this application are described based on the same concept and thus present commonality among them, they can be performed independently unless otherwise specified. Similarly, the various embodiments disclosed in this application are all based on the same inventive concept; therefore, concepts expressed in the same way, as well as concepts that are appropriately changed for convenience but are expressed differently, should be understood equivalently.

[0062] Unless otherwise expressly stated, the various embodiments disclosed in this application can be combined in a cross-cutting manner to flexibly construct new embodiments, as long as such combination does not depart from the inventive spirit of this application and can meet the needs of the prior art or solve a certain deficiency in the prior art. Those skilled in the art should be aware of such modifications.

[0063] Please see Figure 1 as well as Figure 2 In one embodiment of the calibration-free estimation method for the peak power of a power battery of this application, the method includes:

[0064] Step S10: Obtain the first low-power operation data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operation data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model.

[0065] The peak power estimation system for power batteries in terminal devices can obtain the first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model.

[0066] In some embodiments, the first low-power operating data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the source domain battery; the second low-power operating data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the target domain battery; the basic network architecture of the first residual compensation model and the second residual compensation model is a bidirectional long short-term memory network connected in series with a fully connected layer.

[0067] Taking two different models of commercial lithium-ion batteries as an example, one is used as the source domain battery and the other as the target domain battery. The daily low-power operation data uses dynamic stress test (DST) condition data as an example, but is not limited to DST condition.

[0068] In some embodiments, the step of constructing a first fractional-order battery terminal voltage model includes:

[0069] Step S101: Obtain the first low-power operation data of the source domain battery, including battery current, battery state of charge, measured terminal voltage, and battery temperature.

[0070] Step S102: Based on the battery current, battery state of charge, and battery temperature in the first low-power operation data, construct a first fractional-order battery terminal voltage model that uses fractional-order operators to describe the dynamic behavior of the battery terminal voltage. Determine the model parameters of the first fractional-order battery terminal voltage model by minimizing the root mean square error between the predicted terminal voltage output by the first fractional-order battery terminal voltage model and the measured terminal voltage in the first low-power operation data.

[0071] Specifically, please refer to Figure 3 Based on the first low-power operation data of the source domain battery, a first fractional-order battery terminal voltage model is constructed, which uses a fractional-order operator to describe the dynamic behavior of the battery terminal voltage. The expression of the first fractional-order battery terminal voltage model is as follows:

[0072]

[0073] in, This represents the predicted terminal voltage output by the first fractional-order battery terminal voltage model at time t.

[0074] Indicates the battery open-circuit voltage; This represents the battery current at time t; Indicates the ohmic internal resistance of the battery; Represents the polarization voltage at time t; This represents a fractional operator of order α, used to accurately describe the dynamic behavior of battery terminal voltage; Represents the polarization resistance of the polarization branch; Represents the polarization capacitor of the polarization branch; This represents the state of charge of the battery at time t; This represents the battery temperature at time t.

[0075] As can be seen from the above first fractional-order battery terminal voltage model, through the fractional-order operator Characterizing the slow dynamics of battery polarization provides a basic physical model for subsequent residual compensation.

[0076] Furthermore, based on the first low-power operation data, the model parameters of the first fractional-order battery terminal voltage model are determined by minimizing the root mean square error between the predicted and measured terminal voltage values. The objective function for the root mean square error between the predicted and measured terminal voltage values ​​is expressed as:

[0077]

[0078] in, This represents the set of model parameters for the first fractional-order battery terminal voltage model, including the battery open-circuit voltage. Battery internal resistance (ohms) Polarization resistor of polarization branch Polarization capacitor of polarization branch The order α of fractional operators, etc. This represents the total number of sampling points for the first low-power operation data; This represents the measured terminal voltage value in the first low-power operation data; This represents the predicted terminal voltage output from the first fractional-order battery terminal voltage model.

[0079] By minimizing the root mean square error (RMSE) between the predicted and measured terminal voltage values, the optimal model parameters adapted to the low-power operation of the source domain battery are solved, thus completing the construction of the first fractional-order battery terminal voltage model.

[0080] Furthermore, the battery state of charge (SOC) is updated using the ampere-hour integral method, based on the battery SOC at time t from the first fractional-order battery terminal voltage model. Data sampling period and battery rated capacity The battery state of charge at time t+1 is calculated using the following formula:

[0081]

[0082] in, This represents the state of charge of the battery at time t+1; This represents the state of charge of the battery at time t; Indicates the data sampling period; This indicates the battery's rated capacity.

[0083] Furthermore, to provide training sample labels for the subsequent first residual compensation model, the terminal voltage residual value output by the first fractional-order battery terminal voltage model is defined as follows:

[0084]

[0085] in, This represents the terminal voltage residual value output by the first fractional-order battery terminal voltage model at time t, which is also the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery. This represents the measured terminal voltage of the source domain cell; This represents the predicted terminal voltage of the source domain cell.

[0086] Step S20: Use the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and use the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery as sample labels. Train the first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer until convergence. Based on the second low-power operation data of the target domain battery, fine-tune the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence with a low learning rate to obtain the second residual compensation model that matches the target domain battery.

[0087] First low-power operating data of the source domain battery is obtained. A first fractional-order battery terminal voltage model is constructed based on the first low-power operating data of the source domain battery. After determining the predicted terminal voltage value of the source domain battery based on the first fractional-order battery terminal voltage model, the first low-power operating data and the first high-power operating data of the source domain battery are used as training samples. The terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery is used as the sample label. The first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained until convergence. Based on the second low-power operating data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained until convergence are fine-tuned with a low learning rate to obtain a second residual compensation model that matches the target domain battery.

[0088] In some embodiments, please refer to Figure 4 The first residual compensation model uses a bidirectional long short-term memory network (BiLSTM) with fully connected layers in series to construct a residual mapping function, which is used to fit the variation law of the terminal voltage residual of the source domain battery under low power and high power conditions. The expression of the first residual compensation model is:

[0089]

[0090] in, This represents the predicted terminal voltage residual value output by the first residual compensation model at time t; This represents the predicted terminal voltage of the source domain cell; This represents the residual mapping function consisting of a bidirectional long short-term memory network with cascaded fully connected layers; This represents the battery current at time t; This represents the state of charge of the battery at time t; This represents the battery temperature at time t; This represents the set of network parameters for the first residual compensation model.

[0091] By using a bidirectional long short-term memory (BiLSTM) network structure with fully connected layers, the nonlinear mapping law between the battery current, battery state of charge, battery temperature and the predicted and measured terminal voltage residuals of the source domain battery under low and high power conditions is learned, thus preparing for subsequent target domain migration.

[0092] In some embodiments, the step of training the first residual compensation model includes:

[0093] Step S201: Obtain a training sample set, wherein the training sample set includes multiple training samples and their corresponding sample labels, the training samples include the first low-power operation data and the first high-power operation data of the source domain battery, and the sample labels characterize the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery.

[0094] Step S202: Input multiple training samples and their corresponding sample labels into the residual mapping function constructed by a bidirectional long short-term memory network with a fully connected layer. Iterate the network parameters of the residual mapping function to make the residual mapping function fit the variation law of the terminal voltage residual of the source domain battery under low power and high power conditions, until the first residual compensation model is trained to convergence, so as to determine the first residual compensation model that has been trained to convergence.

[0095] Specifically, a training sample set is obtained, wherein the training samples include first low-power operation data and first high-power operation data of the source domain battery, and the first low-power operation data and the first high-power operation data include the battery state of charge. Battery temperature and battery current The sample label represents the predicted terminal voltage of the source domain cell. Measured terminal voltage The residual value of the terminal voltage between .

[0096] Multiple training samples and their corresponding labels are input into the residual mapping function constructed by the above-mentioned Bidirectional Long Short-Term Memory (BiLSTM) network with cascaded fully connected layers; the network parameters of the residual mapping function are then adjusted. Iterative training is performed to fit the residual mapping function to the variation of the terminal voltage residual of the source domain battery under low-power and high-power operating conditions; this continues until the first residual compensation model is trained to convergence, at which point the converged first residual compensation model is determined. The first residual compensation model can use mean squared error (MSE) as the loss function to minimize the predicted terminal voltage residual value. Residual value of terminal voltage The error between them.

[0097] In a further embodiment, the step of fine-tuning the parameters of the final fully connected layer of the first residual compensation model, which has been trained to convergence, with a low learning rate based on the second low-power operating data of the target domain battery, to obtain a second residual compensation model matching the target domain battery, includes:

[0098] Step S2001: Input the second low-power operating data of the target domain battery into the first residual compensation model that has been trained to convergence;

[0099] Step S2002: Fix the network parameters of the bidirectional long short-term memory network in the first residual compensation model, and only use a low learning rate to iteratively fine-tune the parameters of the terminal fully connected layer of the first residual compensation model, so that the fine-tuned first residual compensation model fits the terminal voltage residual change law under the low power condition of the target domain battery. After the fine-tuning is completed, a second residual compensation model matching the target domain battery is obtained.

[0100] Specifically, to enable the first residual compensation model to reconstruct the high-power error structure and achieve high-power calibration-free operation of the target domain battery without acquiring high-power data, this application performs parameter transfer adaptation on the first residual compensation model. During the transfer process, the model parameters are selectively updated: all network parameters of the bidirectional long short-term memory network (BiLSTM) in the first residual compensation model are fixed, only the terminal fully connected layer is retained as trainable, and a lower learning rate is set for this trainable parameter than that used in the source domain training phase, in order to achieve stable, small-sample fine-tuning and avoid overfitting.

[0101] Taking the fully connected layer at the end as the trainable layer as an example, the transfer adaptation operation (Adapt(·) operation) in its parameter update process is calculated using the following formula, which is expressed as:

[0102]

[0103] in, This represents the parameters of the terminal fully connected layer in the first residual compensation model; This represents the parameters of the fully connected layer before the fine-tuning iteration; This represents the parameters of the fully connected layer after fine-tuning the iteration; This indicates a low learning rate; The mean squared error (MSE) loss function is used to quantify the error between the predicted value of the terminal voltage residual and the actual value of the terminal voltage residual. This represents the true residual value of the terminal voltage of the target domain battery at time t, which is calculated from the second low-power operating data of the target domain battery. That is, the residual value of the terminal voltage between the measured value and the predicted value of the terminal voltage of the target domain battery. This represents the residual mapping function consisting of a bidirectional long short-term memory network with cascaded fully connected layers; This represents the second low-power operating data of the target domain battery, which includes operating characteristics such as battery current, battery state of charge, and battery temperature. This represents the set of network parameters for the first residual compensation model, where the parameters of the BiLSTM layers are fixed, and only the parameters of the fully connected layers are fixed. Participate in iterative updates.

[0104] More specifically, the second low-power operating data of the target domain battery. Input to the first residual compensation model that has been trained to convergence Complete the model input; fix the network parameters of the BiLSTM layers, and only adjust the parameters of the terminal fully connected layers. Perform iterative updates;

[0105] Use a low learning rate The mean squared error (MSE) loss function is minimized using the above calculation formula. The first residual compensation model, after fine-tuning, fits the variation law of the terminal voltage residual under the low power condition of the target domain battery; after fine-tuning, the second residual compensation model matching the target domain battery is obtained.

[0106] Step S30: Construct a second fractional-order battery terminal voltage model based on the second low-power operating data of the target domain battery, and perform residual compensation correction on the terminal voltage prediction value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery.

[0107] Using the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and the residual value of the terminal voltage between the predicted and measured terminal voltage values ​​of the source domain battery as sample labels, the first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained to convergence. Based on the second low-power operation data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence are fine-tuned with a low learning rate to obtain a second residual compensation model that matches the target domain battery. Then, a second fractional-order battery terminal voltage model is constructed based on the second low-power operation data of the target domain battery. Based on the second residual compensation model, the terminal voltage prediction value output by the second fractional-order battery terminal voltage model is corrected by residual compensation to obtain the compensated terminal voltage of the target domain battery.

[0108] In some embodiments, the second fractional-order battery terminal voltage model and the first fractional-order battery terminal voltage model are homogeneous but different physical models of the same type. Their core structures, mathematical expressions, and modeling principles are completely identical. The only differences are in the applicable objects, data sources, and application scenarios. Essentially, they are both electrochemical equivalent circuit models based on fractional-order operators to characterize the slow dynamic characteristics of battery polarization, used to achieve accurate prediction of battery terminal voltage. These will not be elaborated upon here.

[0109] In some embodiments, the step of performing residual compensation correction on the predicted terminal voltage value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery includes:

[0110] Step S301: Input the battery current, battery state of charge and battery temperature from the second low-power operation data of the target domain battery into the second residual compensation model to obtain the terminal voltage residual compensation value of the target domain battery output by the second residual compensation model.

[0111] Step S302: Compensate the predicted terminal voltage value output by the second fractional-order battery terminal voltage model according to the terminal voltage residual compensation value to obtain the compensated terminal voltage of the target domain battery.

[0112] Specifically, the battery current, battery state of charge, and battery temperature from the second low-power operation data of the target domain battery are input into the second residual compensation model, which then outputs the terminal voltage residual compensation value of the target domain battery. Complete the acquisition of the compensation amount; convert the terminal voltage prediction value output by the second fractional-order battery terminal voltage model into the actual value. The residual compensation value of the target domain battery terminal voltage output by the second residual compensation model is compared with the value of the residual compensation value. Compensation calculations are performed to obtain the compensated terminal voltage of the target domain battery. Complete the residual compensation correction.

[0113] Based on the second fractional-order battery terminal voltage model and the second residual compensation model of the target domain battery, the predicted terminal voltage value of the target domain battery is corrected by residual compensation to obtain the compensated terminal voltage of the target domain battery. The calculation formula is expressed as follows:

[0114]

[0115] in, The compensated terminal voltage of the target domain battery. This is the predicted terminal voltage of the target domain battery output by the second fractional-order battery terminal voltage model. This represents the target domain battery terminal voltage residual compensation value output by the second residual compensation model.

[0116] Step S40: Based on the compensated terminal voltage, construct a constraint optimization problem that includes upper and lower limit constraints on the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. Solve the constraint optimization problem to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery.

[0117] A second fractional-order battery terminal voltage model is constructed based on the second low-power operating data of the target domain battery. The terminal voltage prediction value output by the second fractional-order battery terminal voltage model is corrected by residual compensation based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery. Based on the compensated terminal voltage, a constraint optimization problem is constructed that includes upper and lower limit constraints corresponding to the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. The constraint optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery to complete the calibration-free estimation of the peak power of the power battery.

[0118] In some embodiments, the step of solving for the peak power of the target domain battery includes:

[0119] Step S401: Based on the compensated terminal voltage, construct an optimization problem with the maximum sustainable constant power as the optimization objective, and including upper and lower limits of battery current, upper and lower limits of terminal voltage, upper and lower limits of battery state of charge, and upper and lower limits of battery temperature.

[0120] Specifically, based on the compensated terminal voltage, a constrained optimization problem is constructed with the maximum sustainable constant power as the optimization objective, and including upper and lower limits for battery current, terminal voltage, battery state of charge, and battery temperature. The expression for the constrained optimization problem includes:

[0121]

[0122] in, This indicates that the target domain battery is within a preset time window. The maximum sustainable constant power that satisfies all constraints; Indicates candidate constant power; This represents the battery current at time t; This represents the terminal voltage of the target domain battery after residual compensation correction; Indicates the lower limit of battery current; Indicates the upper limit of battery current;

[0123] Indicates the lower limit of the terminal voltage; Indicates the upper limit of the terminal voltage; This represents the state of charge of the battery at time t;

[0124] Indicates the minimum state of charge of the battery; Upper limit of battery state of charge; This represents the battery temperature at time t; Indicates the lower limit of battery temperature; Indicates the upper limit of battery temperature; Indicates the preset time window length.

[0125] Step S402: Solve the constraint optimization problem to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, and use the maximum sustainable constant power as the peak power of the target domain battery.

[0126] Specifically, a method combining univariate search and forward recursive simulation is employed. The terminal voltage, battery state of charge, and battery temperature within a preset time window are predicted through recursive updates, ultimately solving for the maximum sustainable constant power. The terminal voltage can be the compensated terminal voltage of the target domain battery. The calculation formula is recursively derived, the battery state of charge can be recursively derived using the ampere-hour integration method, and the battery temperature is recursively derived using an equivalent heat capacity model based on the Joule heating mechanism. The calculation formula is expressed as follows:

[0127]

[0128] in, Indicates the battery's equivalent specific heat capacity; This represents the battery current at time t; This represents the state of charge of the battery at time t; This represents the battery temperature at time t; Indicates the ohmic internal resistance of the battery; Represents the polarization resistance of the polarization branch; Indicates the sampling period.

[0129] Furthermore, genetic algorithms or other global optimization algorithms can be used to determine the initial current. As search variables, the battery current, compensated terminal voltage, battery state of charge, and battery temperature within a preset time window are calculated recursively. The average power within the time window is used as the evaluation metric to determine the peak power estimation result. The optimization problem, with the maximum sustainable constant power as the optimization objective and constrained by upper and lower limits for battery current, terminal voltage, battery state of charge, and battery temperature, is transformed into an optimization problem with the initial current as the primary factor. A univariate optimization problem with a search variable is represented as:

[0130]

[0131] in, The optimal initial current, which is the initial current that satisfies all constraints and maximizes the average power within the preset time window, is the decision variable for the entire optimization problem. The independent variable that takes the maximum value represents the feasible current range. Inside, find the objective function The initial current at which the maximum value is achieved ; Indicates the lower limit of battery current; Indicates the upper limit of battery current; Indicates the preset time window Average power within; Indicates the preset time window length; Indicates the time step index; This represents the compensated terminal voltage of the target domain battery at time t. This represents the average operation within a time window, ensuring the sustainability of power throughout the entire time window; This represents the compensation back-end voltage at the initial time (t=1); This represents the battery current at the initial moment (t=1); This represents the peak power of the target domain battery, which is the maximum sustainable constant power that satisfies all constraints within a preset time window. It is the final output result of the optimization problem.

[0132] This application improves the accuracy of terminal voltage prediction by residual compensation correction. By combining multi-constraint optimization and transfer learning, it achieves accurate estimation of peak power of target domain batteries without high-power data, solving the problems of traditional methods relying on high-power calibration data and poor model generalization.

[0133] In some embodiments, please refer to Figure 5 The battery peak power estimation method of this application without calibration ( Figure 5 (Right side) demonstrates a significant accuracy advantage in estimating the peak power of the target battery. Compared with traditional methods ( Figure 5 Compared to the method on the left, the battery peak power calibration-free estimation method of this application can effectively reduce the estimation error through transfer learning without the need for high power data of the target battery, thereby avoiding the error fluctuation caused by high power calibration in traditional methods.

[0134] Traditional methods typically rely on a large amount of high-power or peak-power calibration data, which not only significantly increases the testing cycle and cost, but also may lead to accelerated battery aging due to high-power testing. Figure 5 The results show that traditional methods suffer from significantly increased estimation errors in the absence of high-power data calibration. The calibration-free peak power estimation method of this application utilizes transfer learning technology to effectively leverage the error structure information of source-domain batteries under high-power conditions, achieving generalization capability across battery models. Even if the target battery has not undergone high-power calibration, the calibration-free peak power estimation method of this application can still stably and accurately estimate the peak power. Figure 5The results show that the battery peak power estimation method without calibration proposed in this application has a smaller error and its overall accuracy is significantly better than that of traditional methods, further verifying its significant advantages in engineering applications.

[0135] As can be seen from the above embodiments, compared with the prior art, the present application addresses the issue that different models or batches of batteries in the prior art often require repeated high-rate or peak power tests to complete calibration. This process not only increases testing costs and development cycles, but may also accelerate battery aging under high-stress conditions, affecting safety and reliability. The present application has, but is not limited to, the following beneficial effects:

[0136] Firstly, this application uses transfer learning to transfer the model deviation characteristics of the source domain battery under high power conditions to the target domain battery, enabling the target domain battery to achieve accurate peak power estimation without high power or peak power calibration, reducing the dependence on high rate testing and reducing the testing cost and cycle of power batteries.

[0137] Secondly, this application adopts a combination of fractional-order battery terminal voltage model and residual compensation model, which effectively improves the model mismatch problem at high rates and significantly improves the peak power estimation accuracy of power batteries.

[0138] Thirdly, this application achieves the reuse of source domain battery knowledge to target domain battery through transfer adaptation, improving the generalization ability across battery models and batches; it constructs a multi-constraint safety optimization solution mechanism to ensure that the estimation results are within the safe operating boundary, thereby enhancing the system's reliability and stability; at the same time, it avoids high-stress condition testing of the target power battery, slows down battery aging, and extends its service life, which has extremely strong engineering practical value.

[0139] Please see Figure 6A calibration-free peak power estimation device for a power battery, provided to meet one of the purposes of this application, includes a terminal voltage prediction module 1100, a compensation model training module 1200, a terminal voltage compensation module 1300, and a peak power estimation module 1400. The terminal voltage prediction module 1100 is configured to acquire first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted terminal voltage value of the source domain battery based on the first fractional-order battery terminal voltage model. The compensation model training module 1200 is configured to use the first low-power operating data and the first high-power operating data of the source domain battery as training samples, use the terminal voltage residual value between the predicted and measured terminal voltage values ​​of the source domain battery as sample labels, and train a first residual compensation model constructed by a bidirectional long short-term memory network with a series fully connected layer until convergence. Based on the second low-power operating data of the target domain battery, the parameters of the final fully connected layer of the trained and converged first residual compensation model are fine-tuned with a low learning rate to obtain a matching target... The system includes a second residual compensation model for the target domain battery; a terminal voltage compensation module 1300, configured to construct a second fractional-order battery terminal voltage model based on the second low-power operating data of the target domain battery, and perform residual compensation correction on the terminal voltage prediction value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery; and a peak power estimation module 1400, configured to construct a constraint optimization problem based on the compensated terminal voltage, including upper and lower limit constraints corresponding to the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery, solve the constraint optimization problem, and obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery to complete the calibration-free estimation of the peak power of the power battery.

[0140] Based on any embodiment of this application, please refer to Figure 7 Another embodiment of this application also provides an electronic device, which can be implemented by a computer device, such as... Figure 7The diagram shows the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable storage medium stores an operating system, a database, and computer-readable instructions. The database may store control information sequences. When the computer-readable instructions are executed by the processor, the processor can implement a calibration-free method for estimating the peak power of a power battery. The processor of the computer device provides computing and control capabilities, supporting the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the calibration-free method for estimating the peak power of a power battery of this application. The network interface of the computer device is used for communication with a terminal. Those skilled in the art will understand that… Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0141] In this embodiment, the processor is used to execute... Figure 6 The specific functions of each module are described, and the memory stores the program code and various data required to execute these modules. The network interface is used for data transmission between the user terminal and the server. In this embodiment, the memory stores the program code and data required to execute all modules in the power battery peak power calibration-free estimation device of this application, and the server can call the server's program code and data to execute the functions of all modules.

[0142] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the peak power estimation method for power batteries without calibration described in any embodiment of this application.

[0143] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the power battery peak power calibration-free estimation method described in any embodiment of this application.

[0144] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0145] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for estimating the peak power of a power battery without calibration, characterized in that, include: Obtain the first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model. Using the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery as sample labels, the first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained to convergence; based on the second low-power operation data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence are fine-tuned with a low learning rate to obtain the second residual compensation model that matches the target domain battery; A second fractional-order battery terminal voltage model is constructed based on the second low-power operating data of the target domain battery. The terminal voltage prediction value output by the second fractional-order battery terminal voltage model is corrected by residual compensation based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery. Based on the compensated terminal voltage, a constraint optimization problem is constructed, which includes upper and lower limit constraints on the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. The constraint optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery.

2. The method for estimating the peak power of a power battery without calibration according to claim 1, characterized in that, The steps for constructing a first-order fractional-order battery terminal voltage model include: Acquire the first low-power operating data of the source domain battery, including battery current, battery state of charge, measured terminal voltage, and battery temperature; Based on the battery current, battery state of charge, and battery temperature in the first low-power operation data, a first fractional-order battery terminal voltage model is constructed that uses fractional-order operators to describe the dynamic behavior of the battery terminal voltage. The model parameters of the first fractional-order battery terminal voltage model are determined by minimizing the root mean square error between the predicted terminal voltage output by the first fractional-order battery terminal voltage model and the measured terminal voltage in the first low-power operation data.

3. The method for estimating the peak power of a power battery without calibration according to claim 1, characterized in that, The steps for training the first residual compensation model include: Obtain a training sample set, wherein the training sample set includes multiple training samples and their corresponding sample labels, the training samples include the first low-power operation data and the first high-power operation data of the source domain battery, and the sample labels characterize the terminal voltage residual value between the predicted terminal voltage value and the measured terminal voltage value of the source domain battery. Multiple training samples and their corresponding sample labels are input into a residual mapping function constructed by a bidirectional long short-term memory network with a fully connected layer. The network parameters of the residual mapping function are iteratively trained to fit the residual mapping function to fit the variation law of the terminal voltage residual of the source domain battery under low power and high power conditions. This process continues until the first residual compensation model is trained to convergence, thus determining the first residual compensation model that has been trained to convergence.

4. The method for estimating the peak power of a power battery without calibration according to claim 1, characterized in that, Based on the second low-power operating data of the target domain battery, the steps of fine-tuning the parameters of the final fully connected layer of the first residual compensation model, which has been trained to convergence, with a low learning rate to obtain the second residual compensation model matching the target domain battery include: The second low-power operating data of the target domain battery is input into the first residual compensation model that has been trained to convergence; The network parameters of the bidirectional long short-term memory network in the first residual compensation model are fixed, and only the parameters of the terminal fully connected layer of the first residual compensation model are iteratively fine-tuned with a low learning rate. This allows the fine-tuned first residual compensation model to fit the terminal voltage residual variation law under the low power condition of the target domain battery. After the fine-tuning is completed, a second residual compensation model matching the target domain battery is obtained.

5. The method for estimating the peak power of a power battery without calibration according to claim 1, characterized in that, The step of performing residual compensation correction on the predicted terminal voltage value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery includes: The battery current, battery state of charge, and battery temperature from the second low-power operating data of the target domain battery are input into the second residual compensation model to obtain the terminal voltage residual compensation value of the target domain battery output by the second residual compensation model. The terminal voltage prediction value output by the second fractional-order battery terminal voltage model is compensated based on the terminal voltage residual compensation value to obtain the compensated terminal voltage of the target domain battery.

6. The method for estimating the peak power of a power battery without calibration according to claim 1, characterized in that, The steps for determining the peak power of the target domain battery include: Based on the compensated terminal voltage, a constrained optimization problem is constructed with the maximum sustainable constant power as the optimization objective, and including upper and lower limits of battery current, terminal voltage, battery state of charge, and battery temperature. The constrained optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window. The maximum sustainable constant power is then used as the peak power of the target domain battery.

7. The method for estimating the peak power of a power battery without calibration according to any one of claims 1 to 6, characterized in that, The first low-power operation data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the source domain battery; the second low-power operation data includes the battery current, battery state of charge, measured terminal voltage, and battery temperature of the target domain battery. The basic network architecture of the first residual compensation model and the second residual compensation model is a bidirectional long short-term memory network connected in series with a fully connected layer.

8. A calibration-free estimation device for the peak power of a power battery, characterized in that, include: The terminal voltage prediction module is configured to acquire the first low-power operating data of the source domain battery, construct a first fractional-order battery terminal voltage model based on the first low-power operating data of the source domain battery, and determine the predicted value of the terminal voltage of the source domain battery based on the first fractional-order battery terminal voltage model. The compensation model training module is configured to use the first low-power operation data and the first high-power operation data of the source domain battery as training samples, and the terminal voltage residual value between the predicted value and the measured value of the terminal voltage of the source domain battery as the sample label. The first residual compensation model constructed by the bidirectional long short-term memory network with a series fully connected layer is trained to convergence. Based on the second low-power operation data of the target domain battery, the parameters of the final fully connected layer of the first residual compensation model that has been trained to convergence are fine-tuned with a low learning rate to obtain the second residual compensation model that matches the target domain battery. The terminal voltage compensation module is configured to construct a second fractional-order battery terminal voltage model based on the second low-power operating data of the target domain battery, and perform residual compensation correction on the terminal voltage prediction value output by the second fractional-order battery terminal voltage model based on the second residual compensation model to obtain the compensated terminal voltage of the target domain battery. The peak power estimation module is configured to construct a constraint optimization problem based on the compensated terminal voltage, which includes upper and lower limit constraints on the battery current, terminal voltage, battery state of charge, and battery temperature of the target domain battery. The constraint optimization problem is solved to obtain the maximum sustainable constant power of the target domain battery that satisfies all constraints within a preset time window, which is taken as the peak power of the target domain battery, so as to complete the calibration-free estimation of the peak power of the power battery.

9. An electronic device comprising a central processing unit and a memory, characterized in that, The central processing unit is used to invoke and run a computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 7, which, when invoked by a computer, executes the steps included in the corresponding method.

Images