Method and system for generating rapid charging strategy of lithium ion battery

Abstract

The invention relates to a rapid charging strategy generation method and system for a lithium ion battery. The method comprises the following steps: determining the model of the lithium ion battery and acquiring corresponding electrochemical parameters; carrying out simulation modeling on the battery and carrying out state initialization; a rapid charging problem is constructed by taking improvement of the charging speed of the battery on the premise of meeting the constraint condition as a target; the constraint conditions comprise limiting the charging current range and the maximum temperature of the battery, and limiting the side reaction overpotential to be a positive value; solving a fast charging problem by using a DDPG algorithm; training a DDPG algorithm by taking the battery simulation model as a simulation environment; in a complete charging process, generating each small section of charging current through cyclic execution of an algorithm; circularly executing a complete charging process for set times, and ending algorithm training; and outputting the optimal charging strategy of the battery of the model. The method and the system are beneficial to improving the battery charging speed and shortening the battery charging time while ensuring the battery safety, and are simple in calculation and wide in application range.

Description

Technical Field

[0001] The present invention relates to the technical field of battery charging, and in particular to a method and system for generating a fast charging strategy for a lithium-ion battery. Background Art

[0002] With the development of the electric vehicle industry, lithium-ion batteries, as the power device of electric vehicles, have received more and more attention in related research. However, long charging time and battery aging are still major challenges facing lithium-ion batteries. Due to the limitations of their own electrochemical characteristics, lithium-ion batteries have the problem of long charging time, which easily causes users' mileage anxiety and seriously restricts the development of the electric vehicle industry. Fast charging usually leads to battery degradation, such as solid electrolyte interface (SEI) growth and lithium deposition, and charging control with excessive current will also bring safety hazards, which may cause fires, explosions and other accidents in severe cases. Therefore, how to strike a balance between fast charging and battery life of lithium-ion batteries is an urgent problem to be solved.

[0003] Traditional battery charging management strategies are mainly divided into model-based methods and model-free methods. Model-based methods use mathematical models to optimize input trajectories, but they have problems such as complex calculations, poor observability, and model uncertainty, so their applications are limited. Model-free methods interact directly with the battery, such as the rule-based adaptive CC-CV method. Although it is widely used due to its simplicity, it is usually too conservative and often lacks optimality, resulting in long charging times. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for generating a fast charging strategy for a lithium-ion battery. The method and system are beneficial to improving the battery charging speed and shortening the battery charging time while ensuring the battery safety, and the calculation is simple and the application range is wide.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a method for generating a fast charging strategy for a lithium-ion battery, comprising the following steps: Step 1: Determine the lithium-ion battery model for which a fast charging strategy needs to be generated, and obtain the corresponding electrochemical parameters of the battery model; Step 2: Simulate and model the battery based on the obtained electrochemical parameters, establish a battery simulation model and initialize the state; Step 3: Construct a fast charging problem with the goal of increasing the charging speed of the battery while satisfying constraints; the constraints include: limiting the charging current range, limiting the maximum battery temperature, and limiting the side reaction overpotential to a positive value; Step 4: Use the DDPG algorithm to solve the fast charging problem; use the battery simulation model established in step 2 as the simulation environment to train the DDPG algorithm in the simulation environment; in a complete charging process, generate each small charging current through the cyclic execution of the DDPG algorithm, thereby realizing a complete charging process; the algorithm training is completed after the complete charging process is cyclically executed for a set number of times; Step 5: Output the optimal charging strategy for this battery model.

[0006] Furthermore, in step 2, the Doyle–Fuller–Newman model is used to simulate the battery modeling; at the same time, the battery aging process is simulated by increasing the internal resistance of the anode film once every set number of complete charging processes.

[0007] Furthermore, in step 2, after the battery simulation model is established, the initial state of the battery is randomly set within a certain range. The initial state includes: the initial temperature of the battery T and battery initialization SoC , the subsequent charging process will simulate the change of battery state based on this initial state.

[0008] Furthermore, in step 3, the fast charging problem is constructed in the form of a Markov decision process, which specifically includes the following steps: 301) Define the state and action space: the action is the continuous charging current, and the state includes the battery voltage, temperature and charge level; in, s t is the state space, a t is the action space, SoC is the charge level of the battery, V is the battery voltage, T is the battery temperature, I is the battery current; 302) Construct constraints, including: limiting the charging current within a set range, limiting the maximum battery temperature, and limiting the side reaction overpotential to a positive value; in, t is the current time point, I ( t ) is the current time t The charging current; Imax It is the maximum limit of charging current, negative value represents charging; T ( t ) is the current time t The battery temperature, T max is the maximum limit of battery temperature; η sr ( t ) is the current time t The side reaction overpotential; 303) Construct a reward function: After selecting the current action, i.e., charging current, the immediate reward of the action is calculated through the reward function to judge whether the current action is good or bad; in, r t+1 As a reward for action, r fast For quick items, r safety ( s t , a t ) is a safety term; the safety term includes the side reaction overpotential term and the temperature term: in, is the side reaction overpotential term, r temp ( s t , a t ) is the temperature term, λ sr is the side reaction overpotential term coefficient, η sr ( t ) is the battery overpotential at the current moment, λ temp is the temperature term coefficient; According to the above formula, when the side reaction overpotential is less than 0 or the battery temperature exceeds the maximum limit, the action is penalized to ensure that the battery is in a safe state during the charging process and to increase the charging speed of the battery.

[0009] Furthermore, in step 4, the DDPG algorithm is trained in a simulation environment, specifically including the following steps: 401) Initialization, including initializing the actor network , Critics Network , experience replay buffer and create target networks, including actor target networks μ' and critics target network Q' ; The structures of the actor-target network and the critic-target network are the same as those of the actor-target network and the critic-target network, respectively, and the parameters of the actor-target network and the critic-target network are copied from the actor-target network and the critic-target network, respectively; 402) interacts with the simulation environment formed by the battery simulation model, and obtains the current battery status from the environment s t , selects a charging current for the battery through the actor target network ;in, is the noise used for exploration; 403) executes the action with the current charging current I Charge for one time step; observe the reward given by the environment r t and the new state s t+1 ;Will Store into the experience replay buffer; 404) Update critic network; 405) Update actor network parameters ; 406) Soft update target network parameters; 407) During a complete charging process, steps 402)-406) are executed cyclically to generate each small section of charging current, thereby realizing a complete charging process; after the complete charging process is executed cyclically for a set number of times, the algorithm training is completed.

[0010] Further, in step 404), updating the critic network specifically includes the following steps: 1) First, randomly sample a small batch of experience from the experience replay buffer ; For each sample, use the target network to calculate the target value y i ; in, is the discount factor, μ' and Q' For the target network; 2) Calculate the predicted value using the current critic network ; 3) Calculate mean square error loss L : in, N is the number of mini-batch samples; 4) Use gradient descent to minimize the loss function and update the critic network parameters , as shown below: in, It means assigning the right part of the formula to the left part. β is the learning rate; Denotes the critic network's parameters The gradient of s i and actions a i The rate of change of the critic network output relative to the change of its parameters.

[0011] Further, in step 405), the actor network parameters are updated , specifically including the following steps: 1) Randomly sample a small batch of experience from the experience replay buffer ; 2) For each state s t , there is action , action value ; Calculate the actor policy gradient: in, is the gradient of the critic network with respect to the action, is the gradient of the actor network with respect to the parameters; 3) Update actor network parameters ; in, α is the learning rate.

[0012] Further, in step 406), the target network parameters are soft updated, and the formula is as follows: in, τ is the small update coefficient set.

[0013] The present invention also provides a lithium-ion battery fast charging strategy generation system, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the above method can be implemented.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1) The present invention takes into account both the charging speed and the battery aging problem in generating a fast charging strategy for lithium-ion batteries. By constructing an optimization problem with the charging speed as the target and the battery aging as the constraint, the battery charging time is minimized while ensuring the battery safety.

[0015] 2) The present invention introduces reinforcement learning (RL) and implements a model-free reinforcement learning method, which does not rely on complex electrochemical models and can learn optimal strategies in interaction with the environment, reducing computational complexity and reducing the impact of model uncertainty through data-driven methods.

[0016] 3) The present invention has strong adaptability and can automatically adapt to parameter changes caused by battery aging, maintain charging efficiency, and can adjust strategies under different usage conditions to extend battery life. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of a method implementation of an embodiment of the present invention; Figure 2 It is a training flow chart of the DDPG algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0019] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0020] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0021] like Figure 1 As shown, this embodiment provides a method for generating a fast charging strategy for a lithium-ion battery, comprising the following steps: Step 1: Determine the lithium-ion battery model for which a fast charging strategy needs to be generated, and obtain the corresponding electrochemical parameters of the battery model.

[0022] Step 2: Simulate and model the battery based on the obtained electrochemical parameters, establish a battery simulation model and initialize the state.

[0023] In this embodiment, the Doyle–Fuller–Newman (DFN) model is used to simulate and model the battery. This model can describe the internal electrochemical process of lithium-ion batteries. By setting the charging current of the battery, the model will output various variables including the state of charge (SoC), voltage, and side reaction overpotential.

[0024] In simulating the aging of the battery, this embodiment increases the internal resistance of the anode film once after each set number of complete charging processes. way to simulate the battery aging process.

[0025] After the battery simulation model is established, the initial state of the battery is randomly set within a certain range. The initial state includes: the initial temperature of the battery T and battery initialization SoC In this embodiment, the initial battery temperature T Take a random value in the range [25, 32]; battery initialization SoC A random value is selected in the range of [0.2, 0.4] to enhance the algorithm exploration. The subsequent charging process will simulate the change of battery state based on this initial state.

[0026] Step 3: Construct a fast charging problem with the goal of increasing the battery charging speed while satisfying constraints, including limiting the charging current range, limiting the maximum battery temperature, and limiting the side reaction overpotential to a positive value.

[0027] In this embodiment, the fast charging problem is constructed in the form of a Markov decision process, which specifically includes the following steps: 301) Define the state and action space: the action is the continuous charging current, and the state includes the battery voltage, temperature and charge level; in, s t is the state space, a t is the action space, SoC is the charge level of the battery, V is the battery voltage, T is the battery temperature, I is the battery current.

[0028] 302) Construct constraints, including: i. Excessive charging current will damage the battery, so limit the charging current within the set range; ii. Excessive temperature can also damage the battery, so the maximum temperature of the battery is limited; iii. The continued presence of lithium deposition will lead to battery damage and aging. Therefore, limiting the side reaction overpotential to a positive value can better avoid this phenomenon. in, t is the current time point, I ( t ) is the current time t The charging current; I max It is the maximum limit of charging current, negative value represents charging; T ( t ) is the current time t The battery temperature, T max is the maximum limit of battery temperature; η sr ( t ) is the current time t The side reaction overpotential.

[0029] 303) Construct a reward function: After selecting the current action, i.e., charging current, the immediate reward of the action is calculated through the reward function to judge whether the current action is good or bad; in, r t+1 As a reward for action, r fast For quick items, r safety ( s t , a t ) is a safety term; the safety term includes the side reaction overpotential term and the temperature term: in, is the side reaction overpotential term, r temp ( s t , a t ) is the temperature term, λ sr is the side reaction overpotential term coefficient, ηsr ( t ) is the battery overpotential at the current moment, λ temp is the temperature term coefficient.

[0030] According to the above formula, when the side reaction overpotential is less than 0 or the battery temperature exceeds the maximum limit, the action is penalized. Through such a design, it is possible to avoid the battery overpotential being less than 0 or the temperature exceeding the maximum limit, thereby ensuring that the battery is in a safe state during the charging process and improving the battery charging speed.

[0031] Step 4: Use the DDPG algorithm to solve the fast charging problem. The DDPG algorithm is a policy gradient-based actor-critic method that is applicable to continuous state and action spaces. A deep neural network is used to parameterize the actor and critic, and the learning process is stabilized by the experience replay buffer.

[0032] The battery simulation model established in step 2 is used as the simulation environment to train the DDPG algorithm in the simulation environment. During a complete charging process, the DDPG algorithm is executed cyclically to generate each small charging current, thereby achieving a complete charging process; after the complete charging process is executed cyclically for a set number of times, the algorithm training is completed.

[0033] like Figure 2 As shown, training the DDPG algorithm in a simulated environment includes the following steps: 401) Initialization, including initializing the actor network , Critics Network , experience replay buffer and create target networks, including actor target networks μ' and critics target network Q' ; The structures of the actor-target network and the critic-target network are the same as those of the actor network and the critic network, respectively, and the parameters of the actor-target network and the critic-target network are copied from the actor network and the critic network, respectively.

[0034] 402) interacts with the simulation environment formed by the battery simulation model, and obtains the current battery status from the environment s t , selects a charging current for the battery through the actor target network ;in, is the noise used for exploration.

[0035] The status you want to obtain The constraints of the action, i.e. the charging current and temperature, are as follows: 403) executes the action with the current charging current I Charge for one time step; observe the reward given by the environment r t and the new state s t+1 ;Will Stored to the experience replay buffer.

[0036] award r t The calculation is done by the following formula. With this reward, both charging speed and battery aging are taken into account.

[0037] in, , , .

[0038] 404) Update the critic network; specifically include the following steps: 1) First, randomly sample a small batch of experience from the experience replay buffer ; For each sample, use the target network to calculate the target value y i ; in, is the discount factor, set to 0.99; μ' and Q' The target network.

[0039] 2) Calculate the predicted value using the current critic network .

[0040] 3) Calculate mean square error loss L : in, N is the number of mini-batch samples, set to 64.

[0041] 4) Use gradient descent to minimize the loss function and update the critic network parameters , as shown below: in, It means assigning the right part of the formula to the left part. βis the learning rate, which determines the distance moved along the gradient direction at each parameter update, and its size affects the efficiency and stability of learning. In this embodiment, β Set to 0.001; Denotes the critic network's parameters The gradient of s i and actions a i The rate of change of the critic network output relative to the change of its parameters.

[0042] 405) Update actor network parameters ; Specifically including the following steps: 1) Randomly sample a small batch of experience from the experience replay buffer .

[0043] 2) For each state s t , there is action , action value ; Calculate the actor policy gradient: in, is the gradient of the critic network with respect to the action, is the gradient of the actor network with respect to the parameters.

[0044] 3) Update actor network parameters ; in, α is the learning rate, set to 0.0001.

[0045] 406) Soft update of target network parameters. The purpose of soft update is to make the parameter changes of the target network smoother, avoid the adverse effects of frequent updates of target values, and prevent large shocks and instability in the algorithm during the learning process. The formula for soft update of target network parameters is as follows: in, τ is a small update coefficient set to 0.001.

[0046] 407) In a complete charging process, steps 402)-406) are executed cyclically to generate each small section of charging current, thereby achieving a complete charging process. In this embodiment, each charging is completed when the battery SoC=0.8.

[0047] The algorithm training is completed after the charging process is executed for the set number of times.

[0048] Step 5: Output the optimal charging strategy for this battery model.

[0049] This embodiment also provides a lithium-ion battery fast charging strategy generation system, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the above method can be implemented.

[0050] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0051] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0052] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps for the functions specified in one or more boxes.

[0054] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for generating a fast charging strategy for a lithium-ion battery, characterized in that: The following steps are involved: Step 1: Determine the lithium-ion battery model for which a fast charging strategy needs to be generated, and obtain the corresponding electrochemical parameters of the battery model; Step 2: Simulate and model the battery based on the obtained electrochemical parameters, establish a battery simulation model and initialize the state; Step 3: Construct a fast charging problem with the goal of increasing the charging speed of the battery while satisfying constraints; the constraints include: limiting the charging current range, limiting the maximum battery temperature, and limiting the side reaction overpotential to a positive value; Step 4: Use the DDPG algorithm to solve the fast charging problem; use the battery simulation model established in step 2 as the simulation environment to train the DDPG algorithm in the simulation environment; in a complete charging process, generate each small charging current through the cyclic execution of the DDPG algorithm, thereby realizing a complete charging process; the algorithm training is completed after the complete charging process is cyclically executed for a set number of times; Step 5: Output the optimal charging strategy for this battery model.

2. A method for generating a fast charging strategy for a lithium-ion battery according to claim 1, characterized in that: In step 2, the Doyle–Fuller–Newman model is used to simulate the battery. At the same time, the battery aging process is simulated by increasing the internal resistance of the anode film once every set number of complete charging processes.

3. A method for generating a fast charging strategy for a lithium-ion battery according to claim 1, characterized in that: In step 2, after the battery simulation model is established, the initial state of the battery is randomly set within a certain range. The initial state includes: the initial temperature of the battery T and battery initialization SoC , the subsequent charging process will simulate the change of battery state based on this initial state.

4. A method for generating a fast charging strategy for a lithium-ion battery according to claim 1, characterized in that: In step 3, the fast charging problem is constructed in the form of a Markov decision process, which specifically includes the following steps: 301) Define the state and action space: the action is the continuous charging current, and the state includes the battery voltage, temperature and charge level; in, s t is the state space, a t is the action space, SoC is the charge level of the battery, V is the battery voltage, T is the battery temperature, I is the battery current; 302) Construct constraints, including: limiting the charging current within a set range, limiting the maximum battery temperature, and limiting the side reaction overpotential to a positive value; in, t is the current time point, I ( t ) is the current time t The charging current; I max It is the maximum limit of charging current, negative value represents charging; T ( t ) is the current time t The battery temperature, T max is the maximum limit of battery temperature; η sr ( t ) is the current time t The side reaction overpotential; 303) Construct a reward function: After selecting the current action, i.e., charging current, the immediate reward of the action is calculated through the reward function to judge whether the current action is good or bad; in, r t+1 As a reward for action, r fast For quick items, r safety ( s t , a t ) is a safety term; the safety term includes the side reaction overpotential term and the temperature term: in, is the side reaction overpotential term, r temp ( s t , a t ) is the temperature term, λ sr is the side reaction overpotential term coefficient, η sr ( t ) is the battery overpotential at the current moment, λ temp is the temperature term coefficient; According to the above formula, when the side reaction overpotential is less than 0 or the battery temperature exceeds the maximum limit, the action is penalized to ensure that the battery is in a safe state during the charging process and to increase the charging speed of the battery.

5. A method for generating a fast charging strategy for a lithium-ion battery according to claim 1, characterized in that: In step 4, the DDPG algorithm is trained in a simulated environment, which specifically includes the following steps: 401) Initialization, including initializing the actor network , Critics Network , experience replay buffer and create target networks, including actor target networks μ' and critics target network Q' ; The structures of the actor-target network and the critic-target network are the same as those of the actor-target network and the critic-target network, respectively, and the parameters of the actor-target network and the critic-target network are copied from the actor-target network and the critic-target network, respectively; 402) interacts with the simulation environment formed by the battery simulation model, and obtains the current battery status from the environment s t , selects a charging current for the battery through the actor target network ;in, is the noise used for exploration; 403) executes the action with the current charging current I Charge for one time step; observe the reward given by the environment r t and the new state s t+1 ;Will Store into the experience replay buffer; 404) Update critic network; 405) Update actor network parameters ; 406) Soft update target network parameters; 407) During a complete charging process, steps 402)-406) are executed cyclically to generate each small section of charging current, thereby realizing a complete charging process; after the complete charging process is executed cyclically for a set number of times, the algorithm training is completed.

6. A method for generating a fast charging strategy for a lithium-ion battery according to claim 5, characterized in that: In step 404), the critic network is updated, which specifically includes the following steps: 1) First, randomly sample a small batch of experience from the experience replay buffer ; For each sample, use the target network to calculate the target value y i ; in, is the discount factor, μ' and Q' For the target network; 2) Calculate the predicted value using the current critic network ; 3) Calculate mean square error loss L : in, N is the number of mini-batch samples; 4) Use gradient descent to minimize the loss function and update the critic network parameters , as shown below: in, It means assigning the right part of the formula to the left part. β is the learning rate; Denotes the critic network's parameters The gradient of s i and actions a i The rate of change of the critic network output relative to the change of its parameters.

7. A method for generating a fast charging strategy for a lithium-ion battery according to claim 5, characterized in that: In step 405), update the actor network parameters , specifically including the following steps: 1) Randomly sample a small batch of experience from the experience replay buffer ; 2) For each state s t , there is action , action value ; Calculate the actor policy gradient: in, is the gradient of the critic network with respect to the action, is the gradient of the actor network with respect to the parameters; 3) Update actor network parameters ; in, α is the learning rate.

8. A method for generating a fast charging strategy for a lithium-ion battery according to claim 5, characterized in that: In step 406), the target network parameters are soft updated, and the formula is as follows: in, τ is the small update coefficient set.

9. A lithium-ion battery fast charging strategy generation system, characterized in that: The method comprises a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method according to any one of claims 1 to 8 can be implemented.

Images