Rapid charging control method for lithium ion battery and related equipment

Abstract

The invention provides a lithium ion battery rapid charging control method and related equipment, and relates to the technical field of battery charging control, and the method comprises the steps: building a state vector based on an acquired external measurable signal and an internal state of a target lithium ion battery; simulating the target lithium ion battery through a constant-current charging experiment or a high-fidelity single-particle model, recording the charging cut-off voltage when the target lithium ion battery reaches the target charge state in different health states, and obtaining the maximum allowable charging voltage; designing a multi-target reward function by taking the maximum charging voltage as a constraint condition, and calculating a feedback value under the current state vector according to the multi-target reward function; inputting the state vector into a reinforcement learning agent to output continuous actions, and optimizing the reinforcement learning agent by using a feedback value under the current state vector to obtain an optimal charging control strategy of the target lithium ion battery; and rapid charging is realized on the premise of ensuring safety and prolonging the service life.

Description

Technical Field

[0001] This invention relates to the field of battery fast charging technology, and in particular to a lithium-ion battery fast charging control method and related equipment. Background Technology

[0002] With the widespread application of electric vehicles (EVs) and energy storage systems, lithium-ion batteries have become the mainstream energy storage unit due to their high energy density, long cycle life, and excellent power performance. However, in fast-charging scenarios, traditional charging methods struggle to balance charging speed and battery life. Long-term high-rate charging can lead to serious side reactions and performance degradation, such as lithium dendrite precipitation, solid electrolyte interface (SEI) film growth, increased polarization, and capacity decay, severely impacting system safety and lifespan.

[0003] Currently, commonly used lithium-ion battery charging methods mainly include constant current-constant voltage (CC-CV), model predictive control (MPC), and data-driven reinforcement learning (RL). The CC-CV method is widely used due to its simple structure and ease of implementation, but its voltage upper limit is fixed and cannot be dynamically adjusted according to the battery's state of health (SoH). As the battery ages, this method continues to perform the same high-voltage constant voltage phase, easily leading to side reactions and thermal stress accumulation, thus accelerating capacity decay and increasing safety risks. The MPC method predicts the future battery state by establishing an electrochemical model, achieving constrained optimization of the charging process. It can balance charging speed and lifespan to some extent, but this method is highly dependent on model integrity and parameter identification. The applied battery parameters change significantly with aging, temperature, and other factors, leading to increased prediction errors. Furthermore, the MPC algorithm needs to solve the optimization problem in real time within each sampling period, resulting in high computational load and hardware requirements, making it difficult to implement in vehicle battery management systems (BMS). Real-time deployment in the system; In recent years, reinforcement learning methods have been introduced into the field of battery charging control. These methods continuously adjust the control strategy according to the reward function through the interaction between the agent and the battery environment, thereby achieving adaptive optimization without the need for an accurate model. However, existing reinforcement learning charging methods still have several problems: First, the training process is not stable enough, and the policy update is prone to oscillation or convergence instability; second, the generalization ability is limited, and the policy performance is difficult to maintain consistency under different battery types, temperatures or aging stages; third, there is a lack of explicit health constraint mechanisms, usually only aiming to shorten charging time or improve energy efficiency, without fully considering the suppression of side reactions or the extension of lifespan; fourth, the algorithm implementation is complex, requiring a large amount of offline simulation or experimental data, making it difficult to directly embed into actual vehicle battery management systems.

[0004] Furthermore, the concept of "side reaction overpotential-SoH mapping constraint" proposed in recent years has provided a new research approach for health-sensing fast charging. This method determines the maximum safe termination voltage of the battery under different health states through experiments or simulations, establishes a mapping relationship between the upper limit of voltage and SoH, and dynamically adjusts the voltage threshold according to the real-time SoH during charging, thereby achieving a balance between charging speed and battery life degradation. However, most existing mapping relationships are based on offline experiments or fixed models and are static constraints that cannot be dynamically updated with real-time changes in battery health. At the same time, such health constraints usually exist independently of the control strategy and do not form a closed-loop collaboration with reinforcement learning or optimization control algorithms, resulting in a gradual decrease in constraint accuracy during long-term operation and weakening the strategy's adaptive ability to changes in battery life. Summary of the Invention

[0005] This invention provides a method and related equipment for controlling fast charging of lithium-ion batteries, the purpose of which is to achieve fast charging while ensuring safety and extending battery life.

[0006] To achieve the above objectives, the present invention provides a method for controlling fast charging of a lithium-ion battery, comprising:

[0007] Step 1: Construct a state vector based on the external measurable signals and internal states of the target lithium-ion battery. The external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health.

[0008] Step 2: Simulate the target lithium-ion battery using a constant current charging experiment or a high-fidelity single-particle model, record the charging cutoff voltage when the target lithium-ion battery reaches the target state of charge under different health conditions, and obtain the maximum allowable charging voltage.

[0009] Step 3: Design a multi-objective reward function with the maximum charging voltage as a constraint, and calculate the feedback value under the current state vector based on the multi-objective reward function;

[0010] Step 4: Input the state vector into the reinforcement learning agent to output continuous actions, and use the feedback value under the current state vector to optimize the reinforcement learning agent to obtain the optimal charging control strategy for the target lithium-ion battery.

[0011] Furthermore, the functional expression for the maximum allowable charging voltage is:

[0012] ;

[0013] in, Indicates a healthy state The maximum permissible charging voltage is as follows: , , , , All are fitting coefficients. This indicates the health status of the target lithium-ion battery.

[0014] Furthermore, the multi-objective reward function is:

[0015] ;

[0016] in, In the state vector The feedback value below, In the state vector The charging speed bonus item below, In the state vector The following health restriction penalties In the state vector The current smoothing term below.

[0017] Furthermore, in the state vector The following are the health constraint penalties:

[0018] ;

[0019] in, This represents the weight of the health constraint. This indicates the terminal voltage of the target lithium-ion battery. Indicates a healthy state The maximum permissible charging voltage.

[0020] Furthermore, the objective function for optimizing the reinforcement learning agent using the feedback value under the current state vector is:

[0021] ;

[0022] in, Represents the objective function value. In the state vector Expectations Represents the dominance function. Indicates the clipping parameters. This indicates a comparison between the old and new strategy concepts.

[0023] Furthermore, the termination condition for optimizing a reinforcement learning agent is:

[0024] The target lithium-ion battery's state of charge reaches the set target value;

[0025] The terminal voltage of the target lithium-ion battery reaches or approaches the maximum allowable charging voltage;

[0026] The target lithium-ion battery has reached a safe temperature threshold.

[0027] The present invention also provides a lithium-ion battery fast charging control device, comprising:

[0028] The acquisition module is used to construct a state vector based on the acquired external measurable signals and internal states of the target lithium-ion battery. The external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health.

[0029] The simulation module is used to simulate the target lithium-ion battery through constant current charging experiments or high-fidelity single-particle models, record the charging cutoff voltage when the target lithium-ion battery reaches the target state of charge under different health conditions, and obtain the maximum allowable charging voltage.

[0030] The design module is used to design a multi-objective reward function with the maximum charging voltage as a constraint, and to calculate the feedback value under the current state vector based on the multi-objective reward function.

[0031] The optimization module is used to input the state vector into the reinforcement learning agent to output continuous actions, and to optimize the reinforcement learning agent using the feedback value under the current state vector to obtain the optimal charging control strategy for the target lithium-ion battery.

[0032] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a lithium-ion battery fast charging control method.

[0033] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for controlling fast charging of a lithium-ion battery.

[0034] The above-described solution of the present invention has the following beneficial effects:

[0035] This invention constructs a state vector based on the acquired external measurable signals and internal states of the target lithium-ion battery. It simulates the target lithium-ion battery through constant current charging experiments or high-fidelity single-particle models, recording the charging cutoff voltage at which the target lithium-ion battery reaches the target state of charge under different health states, thus obtaining the maximum allowable charging voltage. The maximum charging voltage is used as a constraint to design a multi-objective reward function, and the feedback value under the current state vector is calculated based on the multi-objective reward function. The state vector is input into a reinforcement learning agent to output continuous actions, and the feedback value under the current state vector is used to optimize the reinforcement learning agent, obtaining the optimal charging control strategy for the target lithium-ion battery. Compared with existing technologies, this invention introduces a dynamic sensing mechanism of SoH (Solar Hazard) during the charging control process, embedding the mapping relationship between SoH and the upper limit of the charging voltage as a health constraint into the reinforcement learning agent, achieving an adaptive balance between charging rate and lifespan degradation, thereby achieving fast charging while ensuring safety and extending lifespan.

[0036] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0038] Figure 2 This is a structural diagram of the lithium-ion battery fast charging control device in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the structure of the terminal device in an embodiment of the present invention. Detailed Implementation

[0040] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] In the description of this invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0042] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0043] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0044] This invention addresses existing problems by providing a method and related equipment for fast charging control of lithium-ion batteries.

[0045] like Figure 1 As shown, an embodiment of the present invention provides a fast charging control method for lithium-ion batteries, comprising:

[0046] Step 1: Construct a state vector based on the external measurable signals and internal states of the target lithium-ion battery. The external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health.

[0047] Step 2: Simulate the target lithium-ion battery using a constant current charging experiment or a high-fidelity single-particle model, record the charging cutoff voltage when the target lithium-ion battery reaches the target state of charge under different health conditions, and obtain the maximum allowable charging voltage.

[0048] Step 3: Design a multi-objective reward function with the maximum charging voltage as a constraint, and calculate the feedback value under the current state vector based on the multi-objective reward function;

[0049] Step 4: Input the state vector into the reinforcement learning agent to output continuous actions, and use the feedback value under the current state vector to optimize the reinforcement learning agent to obtain the optimal charging control strategy for the target lithium-ion battery.

[0050] In this embodiment of the invention, during the charging control process of a lithium-ion battery, the externally measurable signals of the battery mainly include the terminal voltage. Charging current ,temperature Parameters such as internal states include the state of charge and the state of health, which are obtained through estimation or modeling.

[0051] In order to comprehensively reflect the dynamic characteristics of the battery, this embodiment of the invention selects terminal voltage, state of charge and health state to construct a state vector, so as to simultaneously describe the battery's energy level, degree of health degradation and operating safety boundary.

[0052] In this embodiment of the invention, the terminal voltage is one of the directly perceptible state inputs in reinforcement learning, used to evaluate the current charging safety and health constraint boundaries. It is the most direct and observable external measurement of the battery, and its transient response reflects factors such as the electrochemical reaction rate, polarization voltage, and side reaction intensity. It consists of electrode reaction potential, ohmic voltage drop, and polarization potential, and can be expressed as:

[0053] ;

[0054] in, , These are represented as the open-circuit potential functions for the positive and negative terminals, respectively. This represents the ohmic voltage drop between the electrolyte and the tabs. Indicates the charge transfer polarization potential;

[0055] State of charge (SOC) represents the ratio of the current energy storage level to the rated capacity, used to measure the remaining energy of a battery. Its dynamic change follows the coulomb integral formula:

[0056] ;

[0057] in, The nominal capacity is represented by this variable. In order to reduce integration error, this embodiment of the invention combines Kalman filtering in the battery management system to estimate the state of charge, ensuring the real-time accuracy of the state of charge signal. This variable is used at the control level to define the charging termination condition.

[0058] Battery health status represents the ratio of current usable capacity to initial capacity, and is an important indicator of battery aging. The calculation formula is:

[0059] ;

[0060] in, Indicates the currently available capacity. Indicates the initial capacity;

[0061] Because batteries are affected by various aging mechanisms such as side reactions, lithium deposition, and SEI film growth during long-term charge-discharge cycles, their health status gradually decreases over time. In order to reflect this change in charging control, this embodiment of the invention obtains the health status value through a periodically updated formula. The change in health status not only affects the capacity, but also determines the upper limit of the allowable terminal voltage and the reaction kinetic constraints. It is the core variable for realizing health sensing control in this embodiment of the invention.

[0062] In summary, the state vector defined in this embodiment of the invention is as follows:

[0063] ;

[0064] This state vector ensures the observability of the charging process and reflects the dynamic changes caused by aging, providing complete environmental information for reinforcement learning agents.

[0065] During rapid charging, the main aging mechanisms within the battery include negative electrode lithium deposition, SEI film growth, and accelerated reactions due to overheating. Among these, the negative electrode side-reaction overpotential... ( ) is an important physical quantity for measuring lithium deposition risk. When When the voltage is too low, metallic lithium is easily deposited on the surface of the negative electrode, resulting in irreversible capacity loss. According to electrochemical kinetics theory, the side reaction overpotential is defined as:

[0066] ;

[0067] in, The negative solid-phase potential, The electrolyte potential, For the equilibrium potential of the side reaction (approximately 0V in the lithium deposition reaction), when At this time, the side reactions tend to be activated, the lithium metal deposition rate increases rapidly, leading to an increase in SEI layer thickness and loss of active lithium.

[0068] However, it cannot be directly measured in a real BMS. Therefore, this invention establishes [the necessary frameworks] through experimental and high-fidelity modeling stages. With charging cutoff voltage The correspondence between them transforms the overpotential constraint into an observable voltage constraint, thereby achieving indirect regulation of the health state.

[0069] Therefore, embodiments of the present invention introduce SoH– The basic idea behind the mapping constraint mechanism for achieving health-sensing voltage upper limit control is as follows:

[0070] Under different states of health (SoH) of the battery, the charging cutoff voltage when the target SoC (i.e., SOH of 80%) is reached is recorded through constant current charging experiments or high-fidelity single-particle model simulations. As the battery ages, the negative electrode polarization and internal resistance increase, causing the terminal voltage response to gradually change. It exhibits a trend of monotonic change compared to SoH.

[0071] The present invention obtains the functional expression for the maximum allowable charging voltage by fitting experimental data as follows:

[0072] ;

[0073] in, Indicates a healthy state The maximum permissible charging voltage is as follows: , , , , All are fitting coefficients. This function indicates the health status of the target lithium-ion battery. When the SoH decreases, it automatically reduces the maximum allowable charging voltage to reduce the risk of lithium deposition and thermal stress accumulation.

[0074] During the operation phase of this embodiment of the invention, the health status is obtained through real-time estimation or periodic correction. The estimation methods include capacity decay models, empirical models based on charge / discharge energy, or impedance spectroscopy analysis, which are performed at each time step based on the current health status. The maximum allowable charging voltage can be obtained by querying the mapping curve. .

[0075] In this embodiment of the invention, when the terminal voltage near When the reinforcement learning agent receives a penalty signal, it is prompted to reduce the charging current; when the terminal voltage exceeds... When the voltage drops, a strong penalty is immediately triggered and the output is cut off to prevent continuous overvoltage operation; when the SoH decreases, the mapping function synchronously reduces the allowable voltage, and the control strategy naturally adapts to the aging characteristics to maintain long-term safety and efficiency.

[0076] In order to guide the reinforcement learning agent to balance charging speed and battery life during the learning process, this invention proposes a multi-objective reward function comprising three main components:

[0077] ;

[0078] in, In the state vector The feedback value below, In the state vector The charging speed bonus item below, In the state vector The following health restriction penalties In the state vector The current smoothing term below.

[0079] In this embodiment of the invention, in the state vector Charging speed bonus items Used to encourage reinforcement learning agents to charge the battery to the target state of charge as quickly as possible, defined as:

[0080] ;

[0081] in, This represents the weighting coefficient of the charging rate reward item. This represents the target state of charge. When the state of charge (SoC) rapidly approaches the target value, the penalty value decreases, indicating a higher reward.

[0082] To prevent side reactions in the battery under high voltage, this embodiment of the invention introduces a health mapping relationship, that is, a functional mapping between the health state and the maximum allowable charging voltage. When the terminal voltage exceeds the maximum allowable charging voltage, a penalty is applied in the state vector. The following are the health constraint penalties:

[0083] ;

[0084] in, This represents the weight of the health constraint. This indicates the terminal voltage of the target lithium-ion battery. Indicates a healthy state The maximum allowable charging voltage is determined by dynamically adjusting the upper limit of the voltage. In this embodiment of the invention, the penalty threshold is automatically changed at different health stages, thereby achieving coordinated control of lifespan and speed.

[0085] To prevent polarization and thermal stress caused by sudden current changes, embodiments of the present invention incorporate a smoothness control penalty term into the multi-objective reward function:

[0086] ;

[0087] in, This represents the weighting coefficient of the current smoothing term, which constrains the reinforcement learning agent to output a smooth current within a continuous sampling period, ensuring the hardware safety and temperature rise controllability of the charging system.

[0088] In this embodiment of the invention, the reinforcement learning agent adopts an Actor-Critic dual-network architecture; wherein, the input to the policy network (Actor) is a state vector. The output is an action. The network structure consists of two fully connected layers, with ReLU as the activation function. The output layer uses the tanh function and is linearly mapped to the allowable current range. The input to the value network (Critic) is the same state vector. The output is the state value function. It is used to evaluate the merits of the current strategy. The network structure is the same as that of the Actor network but the parameters are independent.

[0089] In this embodiment of the invention, the reinforcement learning agent limits the policy update magnitude by pruning the probability ratio, thereby ensuring a smooth training process. Therefore, the reinforcement learning agent is optimized using the feedback value under the current state vector, and its optimization objective function is:

[0090] ;

[0091] in, Represents the objective function value. In the state vector Expectations Represents the dominance function. Indicates the clipping parameters. This indicates a comparison between the old and new strategy concepts. .

[0092] The optimization objective function ensures that each policy update does not exceed a predetermined step size, thereby preventing policy collapse or performance oscillations.

[0093] In this embodiment of the invention, in order to improve sample utilization, the generalized dominance estimation method is used to calculate the dominance function, which is expressed as:

[0094] ;

[0095] in:

[0096] ;

[0097] The above formula, by introducing a balancing parameter, achieves a trade-off between bias and variance, significantly improving training stability. The value network updates by minimizing the mean squared error.

[0098] ;

[0099] in, This indicates the cumulative return on the discount. This represents the mathematical expectation of the sampling experience. Indicates by parameters The constructed value function approximator.

[0100] In each training cycle of this embodiment of the invention, the reinforcement learning agent interacts with the simulation environment and samples the state vector. Then perform the following steps:

[0101] Calculate the advantage function ;

[0102] Using the pruning objective function Update the Actor network;

[0103] Minimize value loss function Update the Critic network;

[0104] Repeat the above process until the cumulative reward converges.

[0105] Through the above update strategy, the reinforcement learning agent can gradually learn the optimal charging current distribution under different SoH stages, achieving a long-term stable charging control strategy. To ensure safety and controllability, this embodiment of the invention introduces two levels of constraints at the action output layer: current constraints and voltage constraints.

[0106] Current satisfies ,in, Actions exceeding the range are automatically clipped by the battery management system's hardware settings;

[0107] If the charging strategy output by the intelligent agent causes the terminal voltage to exceed The system immediately applies penalties and restricts action updates to avoid lithium deposition and overheating risks.

[0108] In the policy optimization of the reinforcement learning agent in this embodiment of the invention, health constraints and reward functions work together to enable the reinforcement learning agent to automatically learn behaviors that comply with safety boundaries, thereby achieving adaptive control of health perception without relying on a clear physical model.

[0109] Specifically, the optimization process of the reinforcement learning agent in this embodiment of the invention is as follows:

[0110] 1. Load the parameters of the pre-trained reinforcement learning policy network. With value network parameters Set the current limit Upper temperature limit and charging termination conditions;

[0111] 2. Real-time acquisition of battery terminal voltage Current ,temperature The current state vector is calculated using the SoC and SoH estimation algorithm embedded in the battery management system.

[0112] ;

[0113] 3. Calculate the maximum allowable charging voltage based on the current SoH value:

[0114] ;

[0115] 4. Convert the state vector Input the policy network into the reinforcement learning agent, and output a series of actions. And perform trimming in the motion constraint module:

[0116] ;

[0117] Performing this action may result in If so, the intensity of the action will be automatically reduced or a penalty signal will be applied;

[0118] 5. The output current command is transmitted to the charging execution unit, where the DC / DC converter adjusts the current. The battery management system continuously monitors the changes in terminal voltage and temperature, and provides real-time feedback of the charging results to the reinforcement learning agent.

[0119] 6. Calculate instant rewards based on status changes during the charging process:

[0120] ;

[0121] And calculate the cumulative discount return. During offline or periodic online update phases, reinforcement learning agents utilize state vectors. Update the policy network and value network, and optimize the parameters. .

[0122] Specifically, the termination condition for the optimization of a reinforcement learning agent is:

[0123] The target lithium-ion battery's state of charge reaches the set target value;

[0124] The terminal voltage of the target lithium-ion battery reaches or approaches the maximum allowable charging voltage;

[0125] The target lithium-ion battery has reached a safe temperature threshold.

[0126] This invention constructs a state vector based on the acquired external measurable signals and internal states of the target lithium-ion battery. The target lithium-ion battery is simulated using a constant current charging experiment or a high-fidelity single-particle model. The charging cutoff voltage at which the target lithium-ion battery reaches the target state of charge under different health states is recorded, yielding the maximum allowable charging voltage. A multi-objective reward function is designed using the maximum charging voltage as a constraint, and the feedback value under the current state vector is calculated based on the multi-objective reward function. The state vector is input into a reinforcement learning agent to output continuous actions, and the reinforcement learning agent is optimized using the feedback value under the current state vector to obtain the optimal charging control strategy for the target lithium-ion battery. Compared with existing technologies, this invention introduces a dynamic sensing mechanism of SoH (Sort of Health) during the charging control process, embedding the mapping relationship between SoH and the upper limit of the charging voltage as a health constraint into the reinforcement learning agent. This achieves an adaptive balance between charging rate and lifespan degradation, thereby enabling fast charging while ensuring safety and extending lifespan.

[0127] Corresponding to the lithium-ion battery fast charging control method described in the above embodiments, such as Figure 2 As shown, this embodiment of the invention also provides a lithium-ion battery fast charging control device 100, which includes:

[0128] The acquisition module 101 is used to construct a state vector based on the acquired external measurable signals and internal states of the target lithium-ion battery. The external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health.

[0129] The simulation module 102 is used to simulate the target lithium-ion battery through constant current charging experiments or high-fidelity single-particle models, record the charging cut-off voltage when the target lithium-ion battery reaches the target state of charge under different health states, and obtain the maximum allowable charging voltage.

[0130] Design module 103 is used to design a multi-objective reward function with the maximum charging voltage as a constraint, and to calculate the feedback value under the current state vector based on the multi-objective reward function.

[0131] The optimization module 104 is used to input the state vector into the reinforcement learning agent to output continuous actions, and to optimize the reinforcement learning agent using the feedback value under the current state vector to obtain the optimal charging control strategy for the target lithium-ion battery.

[0132] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0133] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0134] This invention also provides a terminal device, such as... Figure 3As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 3 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, it implements the above-described lithium-ion battery fast charging control method.

[0135] The terminal device D10 can be a desktop computer, laptop, handheld computer, server, server cluster, or cloud server, etc. This terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will understand that... Figure 3 This is merely an example of terminal device D10 and does not constitute a limitation on terminal device D10. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0136] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0137] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0138] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0140] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a fast charging control method for a lithium-ion battery.

[0141] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, 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. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0142] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling fast charging of a lithium-ion battery, characterized by, The method comprises the following steps: Step 1, constructing a state vector based on the collected external measurable signals and internal states of the target lithium ion battery, wherein the external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health; Step 2, simulating the target lithium ion battery through a constant current charging experiment or a high-fidelity single particle model, recording the charging cutoff voltage of the target lithium ion battery when reaching a target state of charge under different states of health, and obtaining a maximum allowable charging voltage; Step 3, designing a multi-objective reward function with the maximum charging voltage as a constraint condition, and calculating a feedback value under the current state vector according to the multi-objective reward function; Step 4, inputting the state vector into a reinforcement learning agent to output a continuous action, and optimizing the reinforcement learning agent by using the feedback value under the current state vector to obtain an optimal charging control strategy of the target lithium ion battery.

2. The method of claim 1, wherein the method further comprises: The function expression of the maximum allowable charging voltage is: ； wherein represents the maximum allowed charging voltage in a healthy state of the target lithium-ion battery, , , , , are fitting coefficients, represents the state of health of the target lithium-ion battery.

3. The method of claim 1, wherein the method further comprises: The multi-objective reward function is: ； wherein, represents a feedback value at a state vector , represents a charging rate reward term at a state vector , represents a health constraint penalty term at a state vector , represents a current smoothing term at a state vector .

4. The method of claim 3, wherein the charging rate is increased to the second charging rate when the temperature of the lithium-ion battery is within a second temperature range. The health constraint penalty term under the state vector x is: ； wherein, represents a weight of the health constraint term, represents an end voltage of the target lithium-ion battery, represents a maximum allowed charging voltage at a state of health under the state of health.

5. The method of claim 1, wherein the method further comprises: The optimization objective function for optimizing the reinforcement learning agent by using the feedback value under the current state vector is: ； wherein, denotes the objective function value, denotes the expectation under the state vector denotes the expectation under the state vector denotes the advantage function, denotes the clipping parameter, denotes the new-old policy concept ratio.

6. The method of claim 1, wherein the method further comprises: The termination condition for optimizing the reinforcement learning agent is: The state of charge of the target lithium ion battery reaches a set target value; The terminal voltage of the target lithium ion battery reaches or approaches the maximum allowable charging voltage; The temperature of the target lithium ion battery reaches a safety threshold.

7. A lithium-ion battery fast charging control device, characterized by, The method comprises the following steps: A collection module is configured to construct a state vector based on the collected external measurable signals and internal states of the target lithium ion battery, wherein the external measurable signals include terminal voltage, charging current and temperature, and the internal states include state of charge and state of health; A simulation module is configured to simulate the target lithium ion battery through a constant current charging experiment or a high-fidelity single particle model, record the charging cutoff voltage of the target lithium ion battery when reaching a target state of charge under different states of health, and obtain a maximum allowable charging voltage; A design module is configured to design a multi-objective reward function with the maximum charging voltage as a constraint condition, and calculate a feedback value under the current state vector according to the multi-objective reward function; An optimization module is configured to input the state vector into a reinforcement learning agent to output a continuous action, and optimize the reinforcement learning agent by using the feedback value under the current state vector to obtain an optimal charging control strategy of the target lithium ion battery.

8. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the lithium ion battery fast charging control method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the lithium ion battery fast charging control method according to any one of claims 1 to 6.

Images