Safe optimization and rapid charging control method for lithium battery

Abstract

The invention relates to the technical field of lithium ion batteries, in particular to a lithium ion battery safe optimization rapid charging control method, which comprises the following steps: S1, constructing a lithium ion battery electric heating aging coupling model, determining a model control equation, and establishing an offline charging environment required by reinforcement learning according to the model; s2, initializing a strategy network and two value networks, defining a reinforcement learning state space and an action space, and setting a charging reward function and a cost function; s3, state space variables are updated according to the battery model in the step S1, and the current state, the charging action, the reward value, the cost value and the state at the next moment are added to an experience pool, rapid charging with electric heating perception and health awareness can be achieved, the expression form of a multi-target optimization problem is reconstructed to enhance the safety of a charging strategy, and the method has the advantages of being high in practicability and the like. And meanwhile, an offline training and online deployment mode is adopted, so that the calculation complexity of online charging is reduced, and the method is more efficient in actual charging application.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium ion batteries, and in particular to a lithium battery safety optimization fast charging control method. Background Art

[0002] "Rechargeable" lithium-ion batteries have brought convenience to daily life, and the key to supporting battery rechargeability is charging technology. In order to alleviate the mileage anxiety of electric vehicle users and improve their satisfaction, the improvement of lithium-ion battery charging speed has also become a hot topic in people's vision. Fast charging at an excessively large rate can significantly reduce the charging time, but it violates safety restrictions and may cause battery overheating, shorten battery life, and even cause safety problems. Therefore, charging lithium-ion batteries is a challenging task, and efficient charging strategies play a vital role in optimizing the operation and extending the life of lithium batteries.

[0003] Traditional charging strategies include the constant current constant voltage (CC-CV) charging protocol widely used in industry and its improved version. Since this type of charging method does not consider the feedback of battery status and safety parameters, the charging effect is not ideal. The prior art proposes a model-based feedback charging strategy. The battery model is mainly divided into an electrochemical model and an equivalent circuit model. On this basis, a charging optimization algorithm is designed to solve the multi-objective optimization problem of minimizing the charging time while ensuring that the charging safety constraints are not violated. In view of the high computational complexity, high model dependence, and weak online traceability of traditional optimization algorithms, a deep reinforcement learning algorithm is introduced.

[0004] The Chinese patent with the patent publication number CN112018465A is based on the DDPG algorithm to train the neural network used to decide the battery charging current, so that the battery can meet the charging physical constraints and complete the charging task. However, this method relies on expert experience to design the reward model, and cannot guarantee that the corresponding charging strategy is optimal. In addition, the charging physical constraints are set as soft constraints, and there are safety hazards in the charging process. Therefore, we propose a lithium battery safety optimization fast charging control method to solve the above-mentioned problems. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art, and a lithium battery safety optimization fast charging control method is proposed, which reconstructs the expression of multi-objective optimization problems to enhance the optimality and safety of the charging strategy, and adopts a distributed soft actor critic (DSAC) algorithm and conservative strategy optimization techniques to perform network iterative updates to achieve fast charging with electrothermal perception and health awareness; the method adopts an offline training and online deployment mode to reduce the computational complexity of online charging, and is more efficient in actual charging applications.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a lithium battery safe optimized fast charging control method, comprising the following steps:

[0007] S1. Construct a coupled model of lithium-ion battery electrothermal aging, determine the model control equations, and establish the offline charging environment required for reinforcement learning based on the model;

[0008] S2. Initialize a policy network and two value networks, define the reinforcement learning state space and action space, and set the charging reward function and cost function;

[0009] S3. Update the state space variables according to the battery model in step S1, and add the current state, charging action, reward value, cost value, and next moment state to the experience pool. Iterate the training of the charging strategy network, value evaluation network, and cost evaluation network constructed based on the distributed soft actor-critic algorithm and conservative strategy optimization technique until convergence, and obtain the converged strategy network model;

[0010] S4. Collect the current terminal voltage, surface temperature, and ambient temperature data of the battery in real time, use the state observer to estimate the unmeasurable battery power state and core temperature, input the processed state information into the charging strategy network trained in step S3, and decide on the optimal charging current trajectory.

[0011] Preferably, the lithium-ion battery electrothermal aging coupling model in step S1 includes an equivalent circuit model for capturing the behavior of the battery at different temperatures, a thermodynamic model for characterizing the internal and surface temperatures of the battery, and an aging model for monitoring the health status of the battery.

[0012] Preferably, the control equation derivation process of the lithium-ion battery electrothermal aging coupling model in step S1 is as follows:

[0013] The physical behavior of the battery is described by a temperature-dependent second-order RC equivalent circuit model. According to Kirchhoff's law, the governing equation can be derived as:

[0014]

[0015] V t (t) = V ocv (SoC)-V1(t)-V2(t)-R0(t,T)I(t)

[0016] In the formula, SoC, I, C n 、V t Respectively represent the battery's state of charge, charging current, nominal capacity and terminal voltage; open circuit voltage V ocvis a polynomial function of SoC; V1 and V2 are the voltages of the two resistor-capacitor pairs respectively; the ohmic internal resistance R0, polarization resistance R1, R2 and polarization capacitance C1, C2 are expressed as functions of SoC and temperature; under the assumption of longitudinal uniformity, a two-state thermal model is used to capture the lumped thermal characteristics of the cylindrical battery, and the governing equation of the heat transfer process is:

[0017]

[0018] Where, T s , T c and T f are the battery surface temperature, core temperature and ambient temperature respectively; R s and R c are thermal convection resistance and thermal conduction resistance respectively; C s and C c are the heat capacities of the battery surface and the core, respectively; the heat generation rate Q is given by the Bernardi heat generation model:

[0019] Q(t)=(V ocv (SoC)-V t (t))I(t)+I(t)T m (t)E n (SoC, t)

[0020]

[0021] In the formula, E n is the entropy change in the electrochemical reaction; T m is the average temperature of the battery; in addition, the semi-empirical model of the Arrhenius equation is used to evaluate the battery capacity decay rate, and the charge throughput can be used to quantify the capacity decay at different charging rates:

[0022]

[0023] Where Ah is the total discharge ampere-hour throughput; R g ,z,B,E a , c are the universal gas constant, power law factor, pre-exponential factor, activation energy and charging rate respectively; ΔSoH is the health state decay rate; Δt is the sampling interval.

[0024] Preferably, the implementation process of step S2 is as follows:

[0025] The state space and action space are defined as:

[0026] s t = {SoC(t), V t (t), T c (t)}

[0027] at ={I(t)|I(t)∈[0,I max ]}

[0028] In the formula, I max is the maximum charging current constrained; the charging strategy can be formulated through a multi-objective optimization problem to minimize the charging time required to reach the desired SoC while ensuring that the charging constraints are met; the reward function is set to be only related to the charging speed:

[0029] r t =|SoC tar -SoC(t)|

[0030] Where, SoC tar is the expected SoC, indicating that the charging process is completed; the charging constraints are designed as cost functions respectively:

[0031]

[0032] C soh (t) = |ΔSoH t |

[0033] In the formula, C volt and C temp Used to punish battery overvoltage and overtemperature, while C soh represents the aging of the battery; the total cost function at each step is described as:

[0034] c t =C volt (t)+C temp (t)+C soh (t).

[0035] Preferably, the cost evaluation network implementation process in step S3 is as follows:

[0036] Rewriting the original constrained optimization problem, the Lagrangian dual form of the problem is:

[0037]

[0038] Where λ is the Lagrange multiplier; and are the estimates of reward and cost functions respectively; d is the constraint threshold; ρ π (s) represents the state distribution caused by the strategy π; the underestimation of the cost function in the charging process is used for the training of other state-action pairs, and accumulates into a large deviation in the temporal difference learning, resulting in charging behavior that violates the constraints; the cost evaluation network is constructed using conservative policy optimization techniques to solve the underestimation problem, and the upper confidence limit of the true cost value is used instead of the estimate, that is, the "mean + weighted standard deviation" of the set is calculated, which is expressed as:

[0039]

[0040] Where E represents the use of E bootstrapped networks with the same network structure and independent training process Cost value estimation is performed; k is a conservative parameter; the cost evaluation network parameters are collectively referred to as ψ, and the parameters of each cost sub-network are updated using the time difference error method.

[0041] Preferably, the charging strategy network and the value assessment network implementation process in step S3 are as follows:

[0042] The overestimation of the reward value function in the charging training of the traditional reinforcement learning method will cause the charging speed to decrease, while the value distribution learning method can significantly improve the overestimation problem of the reward value function, thereby reducing the phenomenon of unstable strategy performance during the learning process;

[0043] The training phase is based on the actor-critic network framework, which includes a charging strategy network and a value evaluation network. The input of the charging strategy network is the lithium battery state space variable, and the output is the action value of controlling charging. The value evaluation network is used to evaluate the pros and cons of the current charging strategy. The parameters of the charging strategy network are collectively referred to as φ, and the parameters of the value evaluation network are collectively referred to as θ.

[0044] The Distributed Soft Actor-Critic (DSAC) algorithm assumes that the distribution of rewards is Gaussian, expressed as Learn the mean and variance of the distribution; dynamically adjust the deviation size through the variance of the distribution to achieve the effect of suppressing over-estimation; the value evaluation network gradient can be expressed by the mean-related gradient and the variance-related gradient:

[0045]

[0046] In the formula, the clipping function C p (·) is used to alleviate gradient explosion or disappearance; b is the clipping boundary to alleviate the sensitivity of reward scale; and They are the target minimum distribution network and the target minimum value network respectively; the time difference error is used to update the value evaluation network parameter θ;

[0047] The safe charging problem can be described as a process of solving the optimal strategy π*, which maximizes the long-term reward while limiting the cumulative cost. The objective function is expressed as:

[0048]

[0049] In the formula, the temperature coefficient α gradually decays during the training process; It means that two value distribution networks are used; the charging strategy network is updated using the time difference error to maximize the reward target.

[0050] Preferably, the charging strategy deployment process in step S4 is as follows:

[0051] First, the state observer and the trained charging strategy network π φ Deploy it to the charging controller; set the charging target and constraint threshold; in each sampling period, collect the current terminal voltage, surface temperature, and ambient temperature data of the battery, use the state observer to estimate the battery SoC and core temperature, input the state space variables into the deployed charging strategy network and obtain the output action, denormalize the action value to obtain the expected charging current and input it into the battery environment until the charging process is completed.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] The present invention makes intelligent decisions on charging behavior based on key states, ensures long battery life while balancing battery charging speed and physical constraint satisfaction, reduces the phenomenon of unstable strategy performance during the learning process by rewriting the expression of multi-objective optimization problems and using distributed reinforcement learning algorithms, improves strategy performance after convergence, and improves the online traceability of charging by adopting an offline training and online deployment model. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is an electrothermal coupling model diagram of a cylindrical lithium battery in a lithium battery safety optimization fast charging control method of the present invention;

[0055] Figure 2 A diagram of the training process of a lithium battery charging strategy based on a distributed reinforcement learning algorithm in a lithium battery safety optimization fast charging control method of the present invention;

[0056] Figure 3 This is a comparison effect diagram of the charging strategy proposed by the lithium battery safety optimization fast charging control method of the present invention and the traditional CCCV charging algorithm. DETAILED DESCRIPTION

[0057] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0058] like Figure 1-Figure 3The method for controlling fast charging of lithium batteries with safety optimization is shown. First, the expression of the charging multi-objective optimization equation is rewritten, the solution process of the safety constraint problem is optimized, and a charging strategy network with a maximum reward function and a minimum cost function is obtained through training. Then, the trained charging strategy neural network is used to intelligently control the charging current according to the current state information of the battery, so as to meet the battery charging safety constraints, meet the needs of fast charging, and protect the health of the battery. Specifically, the following steps are included:

[0059] Step 1: Construct a lithium-ion battery electrothermal aging coupling model, determine its control equation, and establish the offline charging environment required for reinforcement learning based on the model.

[0060] (1.1) Battery model

[0061] The present invention adopts Figure 1 The electrothermal aging coupled model shown is used to describe the characteristics of lithium-ion batteries, including an equivalent circuit model for capturing the behavior of the battery at different temperatures, a thermodynamic model for characterizing the internal and surface temperatures of the battery, and an aging model for monitoring the battery health status.

[0062] The physical behavior of the battery is described by a temperature-dependent second-order RC equivalent circuit model. According to Kirchhoff's law, the governing equation can be derived as:

[0063]

[0064] V t (t) = V ocv (SoC)-V1(t)-V2(t)-R0(t,T)I(t)

[0065] In the formula, SoC, I, C n 、V t Respectively represent the battery's state of charge, charging current, nominal capacity and terminal voltage; open circuit voltage V ocv is a polynomial function of SoC; V1 and V2 are the voltages of the two resistor-capacitor pairs respectively; the ohmic internal resistance R0, polarization resistance R1, R2 and polarization capacitance C1, C2 are expressed as functions of SoC and temperature. Under the assumption of longitudinal uniformity, a two-state thermal model is used to capture the lumped thermal characteristics of cylindrical batteries, and the governing equation of the heat transfer process is:

[0066]

[0067] Where, T s , T c and T f are the battery surface temperature, core temperature and ambient temperature respectively; R s and R care thermal convection resistance and thermal conduction resistance respectively; C s and C c are the heat capacities of the battery surface and the core, respectively; the heat generation rate Q is given by the Bernardi heat generation model:

[0068] Q(t)=(V ocv (SoC)-V t (t))I(t)+I(t)T m (t)E n (SoC, t)

[0069]

[0070] In the formula, E n is the entropy change in the electrochemical reaction; T m is the average temperature of the battery. In addition, the semi-empirical model of the Arrhenius equation is used to evaluate the battery capacity decay rate, and the charge throughput can be used to quantify the capacity decay at different charging rates:

[0071]

[0072] Where Ah is the total discharge ampere-hour throughput; R g ,z,B,E a , c are the universal gas constant, power law factor, pre-exponential factor, activation energy and charging rate respectively; ΔSoH is the health state decay rate; Δt is the sampling interval.

[0073] (1.2) Charging environment

[0074] In order to better represent the safe charging problem, a constrained Markov decision process is used to model the reinforcement learning environment, and the tuple is defined as in and Represent the state space and action space respectively; p represents the probability of transitioning from the current state-action pair to the next state; r, γ, c, d are the immediate reward, discount rate, cost and constraint threshold respectively. At each sampling time, the reinforcement learning agent selects an action based on the current state using the strategy of the policy network, and then the environment returns the reward value, cost value and the next state selected by probability. Step 2: Initialize a policy network and two value networks, define the reinforcement learning state space and action space, and set the charging reward function and cost function.

[0075] (2.1) State space and action space

[0076] s t = {SoC(t), V t (t), T c (t)}

[0077] at ={I(t)|I(t)∈[0,I max ]}

[0078] In the formula, and I max This is the maximum charging current with hard constraints.

[0079] (2.2) Charging multi-objective optimization problem

[0080] The charging strategy can be formulated as a multi-objective optimization problem to minimize the charging time required to achieve the desired SoC while ensuring that the charging constraints are met, as follows:

[0081]

[0082] SV t，min <<V t (t)<<V t，max

[0083] T cmin <<T c (t)<<T c,max

[0084] 1) Reward Function

[0085] Set the reward to only be related to charging behaviors that encourage faster charging:

[0086] r t =|SoC tar -SoC(t)|

[0087] Where, SoC tar is the expected SoC, indicating that the charging process is complete;

[0088] 2) Cost Function

[0089] During the battery charging process, the terminal voltage and core temperature of the battery need to be controlled below a reasonable threshold. At the same time, there are also requirements for the battery capacity decay rate. Based on this, the charging constraints are set as follows:

[0090]

[0091] C soh (t) = |ΔSoH t |

[0092] In the formula, C volt and C temp Used to punish battery overvoltage and overtemperature, while C soh represents the aging of the battery. The total cost function is described as:

[0093] c t =C volt (t)+C temp (t)+C soh (t)

[0094] Step 3: Update the state space variables according to the battery model in step 1, and add the current state, charging action, reward value, cost value, and next moment state to the experience pool. Iterate the training of the charging strategy network, value evaluation network, and cost evaluation network constructed based on the distributed soft actor critic algorithm and conservative strategy optimization techniques until convergence, and obtain the converged strategy network model.

[0095] (3.1) Construction of cost assessment network

[0096] First, the expression of the original multi-objective optimization problem is rewritten. The Lagrangian dual form of the problem is:

[0097]

[0098] Where λ is the Lagrange multiplier; and are the estimates of reward and cost functions respectively; d is the constraint threshold; ρ π (s) represents the state distribution caused by policy π.

[0099] The underestimation of the cost function during the charging process is used to train other state-action pairs and accumulates into large deviations in temporal difference learning, resulting in charging behaviors that violate constraints. A cost evaluation network is constructed using conservative policy optimization techniques to solve the underestimation problem. The upper confidence limit of the true cost value is used instead of the estimate, that is, the "mean + weighted standard deviation" of the set is calculated, expressed as:

[0100]

[0101] Where E represents the use of E bootstrapped networks with the same network structure and independent training process Cost value estimation is performed; k is a conservative parameter. The cost evaluation network parameters are collectively referred to as ψ, and the parameters of each cost sub-network are updated using the time difference error method.

[0102] (3.2) Construction of Charging Strategy Network and Value Assessment Network

[0103] The overestimation of the reward value function in charging training of traditional reinforcement learning methods will cause a decrease in charging speed, while the value distribution learning method can significantly improve the overestimation problem of the reward value function, thereby reducing the phenomenon of unstable strategy performance during the learning process.

[0104] The interactive training process of the environment and algorithm for the lithium battery safety optimization fast charging strategy based on the distributed reinforcement learning algorithm is as follows Figure 2 As shown in the figure, the training phase is based on the actor-critic network framework, which includes a charging strategy network and a value evaluation network. The input of the charging strategy network is the lithium battery state space variable, and the output is the action value of controlling charging. The value evaluation network is used to evaluate the pros and cons of the current charging strategy. The charging strategy network parameters are collectively referred to as φ, and the value evaluation network parameters are collectively referred to as θ.

[0105] The Distributed Soft Actor-Critic (DSAC) algorithm assumes that the distribution of rewards is Gaussian, expressed as Learn the mean and variance of the distribution, and dynamically adjust the deviation size through the variance of the distribution to achieve the effect of suppressing over-estimation; the value evaluation network gradient can be expressed by the mean-related gradient and the variance-related gradient:

[0106]

[0107] In the formula, the clipping function C p (·) is used to alleviate gradient explosion or disappearance; b is the clipping boundary to alleviate the sensitivity of reward scale; and They are the target minimum distribution network and the target minimum value network respectively; the time difference error is used to update the value evaluation network parameters θ.

[0108] The safe charging problem can be described as a process of solving the optimal strategy π*, which maximizes the long-term reward while limiting the cumulative cost. The objective function is expressed as:

[0109]

[0110] In the formula, the temperature coefficient α gradually decays during the training process; It means that two value distribution networks are used; the charging strategy network is updated using the time difference error to maximize the reward target.

[0111] Step 4: Collect the current terminal voltage, surface temperature, and ambient temperature data of the battery in real time, use the state observer to estimate the unmeasurable battery power state and core temperature, input the processed state information into the charging strategy network trained in step S3, and decide on the optimal charging current trajectory.

[0112] The deployment process of the charging strategy is:

[0113] First, the state observer and the trained charging strategy network π φDeploy it to the charging controller; set the charging target and constraint threshold; in each sampling period, collect the current terminal voltage, surface temperature, and ambient temperature data of the battery, use the state observer to estimate the battery SoC and core temperature, input the state space variables into the deployed charging strategy network and obtain the output action, denormalize the action value to obtain the expected charging current and input it into the battery environment until the charging process is completed.

[0114] In this embodiment, a battery fast charging control method considering charging safety and health degradation is verified, and a comparative experiment is conducted with the CCCV strategy (including 2C-CCCV, 4C-CCCV, and 6C-CCCV) widely used in industry. The comparison results are as follows: Figure 3 As shown in Table 1, Figure 3 (a) is a comparison chart of charging current rates; Figure 3 (b) is a comparison diagram of battery state of charge; Figure 3 (c) is a comparison diagram of battery terminal voltage; Figure 3 (d) is a comparison chart of battery core temperature.

[0115] Table 1 Comparison of capacity decay under long-term charging cycles

[0116] Charging algorithm DSAC 2C-CCCV 4C-CCCV 6C-CCCV Maximum capacity 98.21％ 98.92％ 97.64％ 95.23％

[0117] The results show that the CCCV strategy has an irreconcilable contradiction between charging speed and constraint satisfaction; although the 2C-CCCV strategy can achieve thermal safety, the charging speed is significantly slower than other strategies; in contrast, when the current rate is increased to 4C, the risk of overheating is introduced. The charging control strategy based on the distributed reinforcement learning algorithm shows a strong ability to balance charging speed and constraints. It limits the core temperature of the lithium battery to below the set 45°C, and the charging time is close to the 4C-CCCV strategy, but has the ability of thermal perception, and the maximum capacity (under 200 cycles) is greater than 97.64% under the 4C-CCCV strategy and 95.23% under the 6C-CCCV strategy, which indicates a longer healthy life.

[0118] The results show that compared with the traditional CC-CV algorithm, the present invention can shorten the charging time while effectively ensuring charging safety and health degradation, thereby reducing users' charging anxiety.

[0119] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A lithium battery safety optimization fast charging control method, characterized in that: The following steps are involved: S1. Construct a coupled model of lithium-ion battery electrothermal aging, determine the model control equations, and establish the offline charging environment required for reinforcement learning based on the model; S2. Initialize a policy network and two value networks, define the reinforcement learning state space and action space, and set the charging reward function and cost function; S3. Update the state space variables according to the battery model in step S1, and add the current state, charging action, reward value, cost value, and next moment state to the experience pool. Iterate the training of the charging strategy network, value evaluation network, and cost evaluation network constructed based on the distributed soft actor-critic algorithm and conservative strategy optimization technique until convergence, and obtain the converged strategy network model; S4. Collect the current terminal voltage, surface temperature, and ambient temperature data of the battery in real time, use the state observer to estimate the unmeasurable battery power state and core temperature, input the processed state information into the charging strategy network trained in step S3, and decide on the optimal charging current trajectory.

2. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The lithium-ion battery electrothermal aging coupling model in step S1 includes an equivalent circuit model for capturing the behavior of the battery at different temperatures, a thermodynamic model for characterizing the internal and surface temperatures of the battery, and an aging model for monitoring the health status of the battery.

3. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The derivation process of the control equation of the lithium-ion battery electrothermal aging coupling model in step S1 is as follows: The physical behavior of the battery is described by a temperature-dependent second-order RC equivalent circuit model. According to Kirchhoff's law, the governing equation can be derived as: V t (t)=V ocv (SoC)-V1(t)-V2(t)-R0(t,T)I(t) In the formula, SoC, I, C n 、V t Respectively represent the battery's state of charge, charging current, nominal capacity and terminal voltage; open circuit voltage V ocv is a polynomial function of SoC; V1 and V2 are the voltages of the two resistor-capacitor pairs respectively; the ohmic internal resistance R0, polarization resistance R1, R2 and polarization capacitance C1, C2 are expressed as functions of SoC and temperature; under the assumption of longitudinal uniformity, a two-state thermal model is used to capture the lumped thermal characteristics of the cylindrical battery, and the governing equation of the heat transfer process is: Where, T s , T c and T f are the battery surface temperature, core temperature and ambient temperature respectively; R s and R c are thermal convection resistance and thermal conduction resistance respectively; C s and C c are the heat capacities of the battery surface and the core, respectively; the heat generation rate Q is given by the Bernardi heat generation model: Q(t)=(V ocv (SoC)-V t (t))I(t)+I(t)T m (t)E n (SoC,t) In the formula, E n is the entropy change in the electrochemical reaction; T m is the average temperature of the battery; in addition, the semi-empirical model of the Arrhenius equation is used to evaluate the battery capacity decay rate, and the charge throughput can be used to quantify the capacity decay at different charging rates: Where Ah is the total discharge ampere-hour throughput; R g ,z,B,E a , c are the universal gas constant, power law factor, pre-exponential factor, activation energy and charging rate respectively; ΔSoH is the health state decay rate; Δt is the sampling interval.

4. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The implementation process of step S2 is as follows: The state space and action space are defined as: s t ={SoC(t),V t (t),T c (t)} a t ={I(t)|I(t)∈[0,I max ]} In the formula, I max is the maximum charging current constrained; the charging strategy can be formulated through a multi-objective optimization problem to minimize the charging time required to reach the desired SoC while ensuring that the charging constraints are met; the reward function is set to be only related to the charging speed: r t =|SoC tar -SoC(t)| Where, SoC tar is the expected SoC, indicating that the charging process is completed; the charging constraints are designed as cost functions respectively: C soh (t)=|ΔSoH t | In the formula, C volt and C temp Used to punish battery overvoltage and overtemperature, while C soh represents the aging of the battery; the total cost function at each step is described as: c t =C volt (t)+C temp (t)+C soh (t)。 5. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The cost evaluation network implementation process in step S3 is as follows: Rewriting the original constrained optimization problem, the Lagrangian dual form of the problem is: Where λ is the Lagrange multiplier; and are the estimates of reward and cost functions respectively; d is the constraint threshold; ρ π (s) represents the state distribution caused by the strategy π; the underestimation of the cost function during the charging process is used for the training of other state-action pairs and accumulates into a large deviation in the temporal difference learning, resulting in charging behavior that violates the constraints; The cost evaluation network is constructed using conservative strategy optimization techniques to solve the underestimation problem. The upper confidence limit of the true cost value is used instead of the estimate, that is, the "mean + weighted standard deviation" of the set is calculated, which is expressed as: Where E represents the use of E bootstrapped networks with the same network structure and independent training process Cost value estimation is performed; k is a conservative parameter; the cost evaluation network parameters are collectively referred to as ψ, and the parameters of each cost sub-network are updated using the time difference error method.

6. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The implementation process of the charging strategy network and the value assessment network in step S3 is as follows: The overestimation of the reward value function in the charging training of the traditional reinforcement learning method will cause the charging speed to decrease, while the value distribution learning method can significantly improve the overestimation problem of the reward value function, thereby reducing the phenomenon of unstable strategy performance during the learning process; The training phase is based on the actor-critic network framework, which includes a charging strategy network and a value evaluation network. The input of the charging strategy network is the lithium battery state space variable, and the output is the action value of controlling charging. The value evaluation network is used to evaluate the pros and cons of the current charging strategy. The parameters of the charging strategy network are collectively referred to as φ, and the parameters of the value evaluation network are collectively referred to as θ. The Distributed Soft Actor-Critic (DSAC) algorithm assumes that the distribution of rewards is Gaussian, expressed as Learn the mean and variance of the distribution; dynamically adjust the deviation size through the variance of the distribution to achieve the effect of suppressing over-estimation; the value evaluation network gradient can be expressed by the mean-related gradient and the variance-related gradient: In the formula, the clipping function C p (·) is used to alleviate gradient explosion or disappearance; b is the clipping boundary to alleviate the sensitivity of reward scale; and They are the target minimum distribution network and the target minimum value network respectively; the time difference error is used to update the value evaluation network parameter θ; The safe charging problem can be described as a process of solving the optimal strategy π*, which maximizes the long-term reward while limiting the cumulative cost. The objective function is expressed as: In the formula, the temperature coefficient α gradually decays during the training process; It means that two value distribution networks are used; the charging strategy network is updated using the time difference error to maximize the reward target.

7. A lithium battery safety optimization fast charging control method according to claim 1, characterized in that: The charging strategy deployment process in step S4 is as follows: First, the state observer and the trained charging strategy network π φ Deploy it to the charging controller; set the charging target and constraint threshold; in each sampling period, collect the current terminal voltage, surface temperature, and ambient temperature data of the battery, use the state observer to estimate the battery SoC and core temperature, input the state space variables into the deployed charging strategy network and obtain the output action, denormalize the action value to obtain the expected charging current and input it into the battery environment until the charging process is completed.

Images