A user preference fast adaptive power battery charging method considering temperature rise constraint, electronic equipment and storage medium

Abstract

The present application relates to a kind of user preference fast adaptive power battery charging method considering temperature rise constraint, electronic equipment and storage medium, method includes: the voltage and temperature data of power battery are acquired, for constructing power battery electric-thermal coupling model;With shorten charging time, reduce energy loss in the process of charging and inhibit battery temperature rise or temperature rise rate as optimization goal, the iteration optimization of different charging current strategy is carried out using reinforcement learning algorithm, and the basic charging strategy under the target user preference condition is acquired;In the process of iteration optimization, the state transition calculation of charging action is carried out using power battery electric-thermal coupling model, and reward value is output based on reward function;When user preference changes, the weight of different target item in reward function is adjusted by user preference weight parameter, and using transfer learning strategy, the basic charging strategy is migrated to different user preference condition and secondary iteration optimization is carried out, and the optimal charging current strategy is acquired.

Description

Technical Field

[0001] This invention relates to the field of power battery management and intelligent control technology, and in particular to a user preference-based fast adaptive power battery charging method, electronic device, and storage medium that takes into account temperature rise constraints. Background Technology

[0002] With increasing attention being paid to fuel consumption and carbon emissions, electric vehicles, with their power batteries at their core, are gradually becoming an important pathway for the transportation sector to achieve a clean energy transition and zero emissions. Lithium-ion batteries, due to their high energy density, long cycle life, and low self-discharge rate, have been widely used in electric vehicle power systems. However, the long charging time of lithium-ion batteries remains one of the key technological bottlenecks hindering the further large-scale deployment of electric vehicles. Therefore, achieving fast charging while ensuring battery safety and lifespan has become a crucial issue that urgently needs to be addressed in the field of power battery management.

[0003] Fast charging of power batteries is essentially a complex optimization problem constrained by multi-physics coupling. During charging, while excessive charging current can shorten charging time, it also leads to rapid heat accumulation inside the battery, increased energy loss, and exacerbated side reactions, thereby causing battery aging and even safety risks. Therefore, fast charging control strategies need to balance charging speed, thermal safety, and energy efficiency, while strictly meeting safety constraints such as overvoltage and overtemperature.

[0004] Existing fast charging control methods for power batteries can be mainly divided into rule-based methods, model-based methods, data-driven methods, and learning-based methods.

[0005] Rule-based methods typically rely on experience to set charging curves, with the most typical being the constant current / constant voltage charging strategy and its multi-stage improvements. These methods are simple in structure and easy to implement, but due to a lack of in-depth characterization of the battery's internal mechanisms and dynamic characteristics, they often require relatively conservative safety constraints, making it difficult to achieve optimal performance under different operating conditions.

[0006] Model-based methods, by establishing a mathematical model of the power battery and combining it with online optimization algorithms to achieve closed-loop control, can improve the optimality and robustness of the charging strategy to some extent. However, these methods are highly dependent on model accuracy and parameter precision, and the internal state of the battery is difficult to measure directly, usually requiring estimation using a state observer, thus introducing additional uncertainties. Furthermore, as battery performance changes over time, fixed model parameters are difficult to maintain applicability in the long term, limiting their adaptive capabilities.

[0007] Data-driven approaches learn the optimal charging strategy directly from experimental or operational data, avoiding reliance on precise models and thus improving flexibility to some extent. However, these methods typically require a large amount of high-quality experimental data, resulting in high experimental costs, long cycles, and the policy performance being highly sensitive to data distribution.

[0008] In recent years, deep reinforcement learning methods have been increasingly adopted in the field of fast-charging control for power batteries due to their advantages, such as not requiring precise models and being able to handle continuous states and action spaces. Reinforcement learning models the charging process as an interaction between an agent and its environment, enabling the battery management system to learn the optimal charging strategy through trial and error. However, most existing reinforcement learning-based charging strategies only focus on safety constraints such as voltage and temperature limits, failing to adequately consider the impact of temperature rise rate on battery structural stability and localized aging. Furthermore, the optimization objectives in reinforcement learning are typically represented in a weighted form, with different weight parameters corresponding to different optimization preferences; however, existing research has not adequately addressed the impact of weight changes on charging results or the policy's adaptability to changes in user preferences.

[0009] Furthermore, when users' preferences for charging speed, safety, or energy efficiency change, traditional reinforcement learning methods often require retraining the strategy, resulting in high training costs and making it difficult to meet the demands for rapid response and flexible adjustment in practical applications. Therefore, how to achieve rapid adaptation of charging strategies to different user preferences while considering temperature rise constraints remains a pressing issue in the field of fast charging control for power batteries. Summary of the Invention

[0010] To address the problems existing in the prior art, the present invention aims to provide a user preference-adaptive power battery charging method, electronic device, and storage medium that considers temperature rise constraints, thereby solving the problems of difficulty in balancing charging speed, temperature rise control, and energy efficiency, as well as the insufficient adaptability of existing strategies to changes in user preferences and high training costs.

[0011] To achieve the above objectives, the present invention provides the following solution: A user preference-based fast adaptive power battery charging method considering temperature rise constraints includes: The voltage and temperature data of the power battery under different charging currents and ambient temperatures are obtained, and the voltage and temperature data are used to construct an electro-thermal coupling model of the power battery. With the optimization objectives of shortening charging time, reducing energy loss during charging, and suppressing battery temperature rise or temperature rise rate, a reinforcement learning algorithm is used to iteratively optimize different charging current strategies to obtain the basic charging strategy under the target user's preference conditions. During the iterative optimization process, the electric-thermal coupling model of the power battery is used to calculate the state transition of the charging action, and the reward value is output based on the reward function. When user preferences change, the weights of different target items in the reward function are adjusted by the user preference weight parameter, and a transfer learning strategy is adopted to transfer the basic charging strategy to different user preference conditions for secondary iterative optimization to obtain the optimal charging current strategy.

[0012] Optionally, acquiring the voltage and temperature data of the power battery under different charging currents and ambient temperatures includes: The power battery was tested for electrical and thermal performance under multiple ambient temperature conditions. The current, voltage and external surface temperature information of the battery during the charging process were collected to obtain voltage change characteristics and temperature change characteristics as the voltage and temperature data.

[0013] Optionally, the power battery electro-thermal coupling model includes: A fractional-order equivalent circuit model is used to describe the dynamic electrical characteristics of the power battery during charging, where the voltage changes with the battery's state of charge and charging current. A one-dimensional lumped parameter thermal model is used to describe the temperature change process of the power battery during charging.

[0014] Optionally, constructing the electro-thermal coupling model of the power battery includes: The heat generation rate is calculated using the fractional-order equivalent circuit model. The heat generation rate is then input into the one-dimensional lumped parameter thermal model to obtain temperature results. These results are used to modify the temperature-related parameters of the fractional-order equivalent circuit model, thereby dynamically coupling electrical and thermal behaviors to generate the power battery electro-thermal coupling model.

[0015] Optionally, the optimization objective includes: ; in, This indicates the cost of charging time. Indicates the cost of temperature rise. Indicates the cost of energy loss. These represent the weights of each cost.

[0016] Optionally, the method further includes: Safety constraints are set for the charging process, including: charging current constraints, upper limit constraints for battery terminal voltage, upper limit constraints for battery temperature, and state of charge range constraints.

[0017] Optionally, the reward function includes: ; in, This indicates the cost of charging time. Indicates the cost of temperature rise. Indicates the cost of energy loss. These represent the weights of each cost. and These represent the over-temperature threshold penalty and the over-voltage threshold penalty, respectively.

[0018] To achieve the above objectives, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements each step of the charging method described in any one of the claims.

[0019] To achieve the above objectives, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of any of the charging methods described herein.

[0020] The beneficial effects of this invention are as follows: I. This invention constructs an electro-thermal coupling model that combines a fractional-order equivalent circuit model with a one-dimensional lumped-parameter thermal model. This model can accurately describe the voltage and temperature changes of a power battery during fast charging over a wide range of ambient temperatures, thereby improving the accuracy and applicability of the charging control model.

[0021] Second, this invention incorporates charging time, charging energy loss, and battery temperature rise into a multi-objective optimization framework, and introduces voltage and temperature safety constraints during the optimization process. This achieves fast charging while ensuring the safe operation of the power battery, taking into account both charging efficiency and thermal safety.

[0022] Third, this invention transforms the problem of fast charging of power batteries into a deep reinforcement learning decision problem in a continuous state space and a continuous action space. It can obtain an optimized charging strategy through interactive learning with the electro-thermal coupling model without the need for an exact analytical model, thereby improving the flexibility and robustness of the control method.

[0023] Fourth, by introducing adjustable weight parameters into the reward function, this invention enables the charging strategy to adaptively adjust according to different users' preferences for charging speed, energy efficiency, and temperature rise control, thereby improving the personalization of the fast charging strategy for power batteries.

[0024] Fifth, this invention adopts a transfer learning strategy, which eliminates the need to retrain the reinforcement learning model from scratch when user preferences change. Only a small amount of retraining is required to complete the strategy adjustment, which significantly reduces training costs and improves the efficiency of recharging strategy updates.

[0025] VI. Through simulation verification and comparison with constant current and constant voltage charging strategies and multi-stage constant current charging strategies based on Bayesian optimization, this invention demonstrates that the proposed fast charging method has superior comprehensive performance in terms of charging time, energy loss, and temperature rise control, and has good prospects for engineering applications. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of a user preference-based fast adaptive power battery charging method considering temperature rise constraints according to an embodiment of the present invention. Figure 2 This is a schematic diagram of a fractional-order electro-thermal coupling model according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the experimental process of an embodiment of the present invention; Figure 4 This is a schematic diagram of the fitting results of the fractional-order model in an embodiment of the present invention; Figure 5 This is a schematic diagram of the fitting results of the thermal model in an embodiment of the present invention; Figure 6 The diagram illustrates the impact of energy loss weights on optimization results in an embodiment of the present invention; (a) shows the change in reward during the training process; (b) shows a comparison of training results in fast charging time; (c) shows a comparison of training results in energy loss; (d) shows a comparison of training results in temperature rise control; (e) shows the change in charging current during the optimized fast charging process; (f) shows the change in SOC during the optimized fast charging process; (g) shows the change in battery voltage during the optimized fast charging process; and (h) shows the change in battery temperature during the optimized fast charging process. Figure 7The diagram illustrates the impact of the temperature rise effect weight on the optimization results in an embodiment of the present invention; (a) shows the reward change during the training process; (b) shows the comparison of training results in fast charging time; (c) shows the comparison of training results in energy loss; (d) shows the comparison of training results in temperature rise control; (e) shows the charging current change during the optimized fast charging process; (f) shows the SOC change during the optimized fast charging process; (g) shows the battery voltage change during the optimized fast charging process; and (h) shows the battery temperature change during the optimized fast charging process. Figure 8 The diagram illustrates the impact of different weight combinations on the optimization results in an embodiment of the present invention; (a) shows the impact of different weight combinations on fast charging time; (b) shows the impact of different weight combinations on energy loss; and (c) shows the impact of different weight combinations on temperature rise. Figure 9 The diagram shows the transfer training results of an embodiment of the present invention; (a) shows the change in reward during the transfer training process; (b) shows the comparison of transfer results in fast charging time; (c) shows the comparison of transfer results in energy loss; and (d) shows the comparison of transfer results in temperature rise control. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] This embodiment discloses a user preference-based fast adaptive power battery charging method that considers temperature rise constraints, including the following steps: Step S100: Conduct electrical and thermal performance tests on the power battery under different ambient temperature conditions, and obtain voltage and temperature data of the battery under different charging currents and ambient temperatures to build an electro-thermal coupling model of the power battery. The electrical and thermal performance tests described in step S100 are charging experiments conducted under multiple ambient temperature conditions, including at least the battery's normal operating temperature. The experimental conditions for identifying the battery's electrical characteristics should include at least open-circuit voltage tests and dynamic operating condition tests at various temperatures. Furthermore, the dynamic operating condition tests should include identification and verification groups to fully verify the identification results of the battery's electrical parameters. The battery's thermal characteristic parameters are determined by charging the power battery under different charging current conditions at room temperature.

[0031] The experiment collects real-time data on the terminal voltage, surface temperature, or internal temperature of the power battery, thereby obtaining the voltage variation characteristics and temperature rise variation law of the power battery under different charging rates and different ambient temperatures, providing basic data support for the subsequent construction of the electro-thermal coupling model.

[0032] Step S200: Based on the test data, identify the electrical and thermal parameters of the power battery and establish an electro-thermal coupling model that can reflect the charging behavior of the power battery in a wide range of ambient temperatures. The electro-thermal coupling model in step S200 includes an electrical model and a thermal model. The electrical model is used to describe the dynamic characteristics of the voltage of the power battery as a function of the state of charge and the charging current during the charging process. The thermal model is used to describe the temperature change process of the power battery caused by ohmic internal resistance loss during the charging process. The electrical sub-model and the thermal model are coupled through the charging current and energy loss to achieve a joint characterization of the electrical and thermal behaviors of the power battery during the charging process.

[0033] Step S300: Based on the state of charge, temperature and charging current constraints of the power battery, construct a multi-objective optimization problem for fast charging with charging time, temperature rise rate and energy loss as optimization objectives; The multi-objective optimization problem for fast charging described in step S300 aims to shorten charging time, suppress battery temperature rise, and reduce energy loss during charging. The constraints of the optimization problem include: the charging current does not exceed a preset maximum charging current threshold, the battery terminal voltage does not exceed the safe cutoff voltage, the battery temperature does not exceed the safe upper temperature limit, and the battery state of charge is within the allowable charging range.

[0034] Step S400: The multi-objective optimization problem is transformed into a reinforcement learning decision problem. The state of charge, temperature and voltage of the power battery are used as state variables, and the charging current is used as action variable. The charging strategy is trained by setting a reward function that includes charging time, temperature rise and energy loss, as well as temperature over-threshold penalty and voltage over-threshold penalty. In the reinforcement learning decision problem described in step S400, the battery's state of charge, temperature, and voltage are used as the state inputs to the reinforcement learning algorithm, and the charging current is used as the action output. The state transition calculation for the charging action is performed using the electro-thermal coupling model, and a reward value is output according to the reward function, thereby enabling the reinforcement learning agent to iteratively train and optimize different charging current strategies.

[0035] Step S500: Introduce user preference weight parameters to adjust the weights of different target items in the reward function, so that the charging strategy can adaptively adjust according to the preferences of different users in terms of charging speed, temperature rise rate and energy efficiency. The reward function in step S500 consists of multiple sub-reward items. These sub-reward items include at least a time reward item related to charging time, a safety reward item related to battery temperature rise, and an efficiency reward item related to charging energy loss. They also include voltage and temperature threshold protection items. By adjusting the weight parameters corresponding to each sub-reward item, the reinforcement learning algorithm can exhibit different optimization tendencies for charging speed, temperature rise rate, and charging efficiency under different user preference conditions, thereby achieving user preference-adaptive fast charging control.

[0036] Step S600: Using a transfer learning strategy, the charging strategy trained under the initial user preference conditions is transferred to different user preference conditions to complete the retraining of the fast charging strategy and achieve rapid adaptation to different user preferences. The transfer learning strategy described in step S600 includes transferring the parameters, policy network, and value function of the pre-trained reinforcement learning model to new user preference conditions, setting a new reward feedback function, and retraining to reduce the training time required for retraining, achieve rapid adaptation to different user preferences, and maintain the stability and security of the fast charging strategy.

[0037] Step S700: Verify the trained charging strategy using simulation experiments, and compare it with the constant current and constant voltage charging strategy and the multi-stage constant current fast charging strategy based on Bayesian optimization to verify the effectiveness and superiority of the proposed optimization strategy.

[0038] In step S700, a simulation charging experiment is conducted using the constructed power battery simulation model. The simulation is compared with the traditional constant current and constant voltage charging strategy and the multi-stage constant current charging strategy based on Bayesian optimization to verify the effectiveness and superiority of the proposed optimization strategy.

[0039] This embodiment provides a user preference-based fast adaptive power battery charging method that considers temperature rise constraints, including: Step S100: Conduct electrical and thermal performance tests on the power battery under different ambient temperature conditions to obtain voltage and temperature data of the power battery under different charging currents and different ambient temperatures; the tests are used to characterize the electrical and thermal responses of the power battery over a wide range of ambient temperatures, providing basic data support for the subsequent construction of an electro-thermal coupling model.

[0040] Step S200: Based on the test data obtained in step S100, the electrical and thermal parameters of the power battery are identified, and an electro-thermal coupling model that can reflect the charging behavior of the power battery in a wide range of ambient temperatures is established; wherein, the electro-thermal coupling model is used to describe the dynamic characteristics of the power battery's terminal voltage changing with the state of charge and charging current during the charging process, as well as the thermal dynamic characteristics of the power battery's temperature changing with energy loss.

[0041] Step S300: Based on the state of charge, temperature, and charging current constraints of the power battery, construct a multi-objective optimization problem for fast charging with charging time, temperature rise rate, and energy loss as optimization objectives; wherein the constraints include at least charging current constraints, battery terminal voltage safety threshold constraints, temperature safety threshold constraints, and state of charge range constraints to ensure the safety and controllability of the fast charging process.

[0042] Step S400: The multi-objective optimization problem described in step S300 is transformed into a reinforcement learning decision problem. The state of charge, temperature, and terminal voltage of the power battery are used as state variables, and the charging current is used as action variables. A joint reward function including charging time, temperature rise rate, and energy loss is constructed, and temperature and voltage over-threshold penalty terms are set to penalize behaviors that violate safety thresholds. The charging strategy is trained in the simulated charging environment constructed by the electro-thermal coupling model to obtain an optimized charging control strategy that meets safety constraints.

[0043] Step S500: Introduce user preference weight parameters to adjust the weights of different objective items in the reward function, so that the charging strategy can adaptively adjust according to the preferences of different users in terms of charging speed, temperature rise rate control and energy efficiency, thereby forming a multi-objective fast charging optimization control strategy for different user preferences.

[0044] Step S600: Using a transfer learning strategy, the charging strategy trained under the initial user preference conditions is transferred to different user preference conditions, and the fast charging strategy is quickly updated through a small amount of retraining, so that the charging strategy can quickly adapt to different user preferences while maintaining safety and stability.

[0045] Step S700: Verify the trained charging strategy using simulation experiments, and compare and analyze it with the constant current and constant voltage charging strategy and the multi-stage constant current fast charging strategy based on Bayesian optimization. Verify the effectiveness and superiority of the proposed optimization strategy from indicators such as charging time, temperature rise rate and energy loss.

[0046] Further, in step S100, the electrical and thermal performance of the power battery are tested under different ambient temperatures. The ambient temperatures at least cover the normal operating temperature range and high-temperature operating range of the power battery, such as 0℃ to 50℃. After completing open-circuit voltage tests and dynamic operating condition tests at different temperatures, thermal characteristic experiments of the battery are conducted. During all tests, the current, voltage, and external surface temperature information of the battery during charging are collected to obtain the voltage and temperature change characteristics of the power battery under different charging rates and ambient temperatures, providing experimental data support for subsequent model construction and parameter identification.

[0047] Further, in step S200, based on the test data obtained in step S100, an electro-thermal coupling model of the power battery is established. The electro-thermal coupling model consists of a fractional-order equivalent circuit model and a one-dimensional lumped-parameter thermal model. The fractional-order equivalent circuit model describes the dynamic electrical characteristics of the power battery during charging, where the voltage changes with the battery's state of charge and charging current. The one-dimensional lumped-parameter thermal model describes the temperature change process of the power battery during charging. The fractional-order equivalent circuit model includes an open-circuit voltage source, an ohmic internal resistance, and two fractional-order RC branches to characterize the polarization effect and memory characteristics of the power battery. Each fractional-order RC branch consists of a resistive element and a constant-phase element, used to describe the battery's polarization behavior. The impedance expression of the constant-phase element is: ; in, These are fractional capacitance parameters. For fractional order, This is the Laplace operator. The model state variables include the voltages of each fractional branch and the battery state of charge. The model parameters change with battery temperature, thus reflecting the influence of temperature on the battery's electrical characteristics. Its state update can be expressed as: ; in, This is the model state vector, containing the voltages of each fractional branch and the battery state of charge. For charging current input, The length of the memory is fractional. (Matrix) , as well as The voltage is determined by the fractional-order parameters, resistance parameters, and sampling time. Based on Kirchhoff's voltage law, the terminal voltage of the power battery is determined by the open-circuit voltage, the fractional-order branch voltage, and the ohmic internal resistance voltage drop, and its expression is: ; in, This refers to the battery terminal voltage. The open-circuit voltage related to the state of charge; The voltage is a fractional-order branch voltage; This is the charging current; The internal resistance is ohms. The open-circuit voltage was obtained through experimental testing, and a mapping relationship between it and the state of charge was established.

[0048] The one-dimensional lumped-parameter thermal model is constructed based on the principle of heat conservation, and describes the heat conduction and convection relationships between the battery interior, surface, and environment through heat capacity and thermal resistance parameters; the discrete update form of the power battery surface temperature can be expressed as: ; in, This refers to the surface temperature of the battery. The ambient temperature; , The coefficient is determined by the battery's thermal resistance and thermal capacity parameters; This refers to the heat generation rate of the battery during the charging process.

[0049] The heat generation term of the thermal model is determined by the energy loss calculated from the fractional-order electrical model, thereby realizing the bidirectional coupling between the electrical model and the thermal model. That is, the electrical behavior affects the heat generation rate, and the temperature change, in turn, affects the electrical model parameters.

[0050] Further, in step S300, a multi-objective optimization problem for fast charging of the power battery is constructed based on the electro-thermal coupling model. The multi-objective optimization problem aims to shorten the charging time, reduce energy loss during the charging process, and suppress the rise in battery temperature or the rate of temperature rise, and can be expressed as: ; in, This indicates the cost of charging time. Indicates the cost of temperature rise. Indicates the cost of energy loss. These represent the weights of each cost.

[0051] Meanwhile, safety constraints are set for the charging process, including at least charging current constraints, upper limit constraints on battery terminal voltage, upper limit constraints on battery temperature, and state of charge range constraints, to ensure the safety and stability of the charging process.

[0052] Further, in step S400, the multi-objective optimization problem described in step S300 is transformed into a Markov decision problem. The state of charge, surface temperature, and terminal voltage of the power battery are used as state variables of the deep reinforcement learning algorithm, and the charging current is used as the action variable of the deep reinforcement learning algorithm. The agent in the deep reinforcement learning selects an action based on the current battery state and performs state transition calculations for the charging action through the electro-thermal coupling model. The model simulation experiment obtains the state information of the next battery. Combined with the reward feedback mechanism, the deep reinforcement learning agent can interact and learn with the battery model in the simulation environment. A reward function is constructed to reflect the optimization effect of the charging strategy. The reward function includes reward terms related to charging time, reward terms related to energy loss, and reward terms related to temperature rise or temperature rise rate. At the same time, penalty terms are introduced for battery terminal voltage exceeding the safety threshold and battery temperature exceeding the safety threshold to suppress charging behavior that violates safety constraints and guide the reinforcement learning agent to learn a charging strategy that meets safety requirements. The constructed reward function is expressed as follows: ; in, , and These represent the over-temperature threshold penalty and the over-voltage threshold penalty, respectively, and their specific expressions are as follows: .

[0053] Furthermore, in step S500, the weights of each objective reward item in the reward function are adjusted according to user preferences, so that the reinforcement learning algorithm exhibits different optimization tendencies for charging time, energy efficiency and temperature rise control under different weight configurations, thereby enabling the trained charging strategy to adapt to the preferences of different users in terms of charging speed, safety and efficiency.

[0054] Furthermore, in step S600, a transfer learning strategy is adopted. While keeping the state space and action space of the reinforcement learning problem unchanged, the parameters of the reinforcement learning model trained under the initial user preference conditions are transferred to the new user preference conditions as initial parameters. After adjusting the weight of the reward function, a small amount of retraining is performed, so that the charging strategy can quickly adapt to the new user preferences with fewer training samples and a shorter training time.

[0055] Furthermore, in step S700, the trained charging strategy is simulated and verified using the constructed electro-thermal coupling model. The charging strategy is then compared and analyzed with the constant current and constant voltage charging strategy and the multi-stage constant current charging strategy based on Bayesian optimization. The effectiveness and superiority of the charging strategy proposed in this invention are verified from multiple dimensions such as charging time, energy loss, and temperature rise control.

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] like Figure 1 As shown, this embodiment discloses a user preference-based fast adaptive power battery charging method considering temperature rise constraints, including: Step S100: Conduct electrical and thermal performance tests on the power battery under different ambient temperature conditions, and obtain voltage and temperature data of the battery under different charging currents and ambient temperatures to build an electro-thermal coupling model of the power battery. Step S200: Based on the test data, identify the electrical and thermal parameters of the power battery and establish an electro-thermal coupling model that can reflect the charging behavior of the power battery in a wide range of ambient temperatures. Step S300: Based on the state of charge, temperature and charging current constraints of the power battery, construct a multi-objective optimization problem for fast charging with charging time, temperature rise rate and energy loss as optimization objectives; Step S400: The multi-objective optimization problem is transformed into a reinforcement learning decision problem. The state of charge, temperature and voltage of the power battery are used as state variables, and the charging current is used as action variable. The charging strategy is trained by setting a reward function that includes charging time, temperature rise and energy loss, as well as temperature over-threshold penalty and voltage over-threshold penalty. Step S500: Introduce user preference weight parameters to adjust the weights of different target items in the reward function, so that the charging strategy can adaptively adjust according to the preferences of different users in terms of charging speed, temperature rise rate and energy efficiency. Step S600: Using a transfer learning strategy, the charging strategy trained under the initial user preference conditions is transferred to different user preference conditions to complete the retraining of the fast charging strategy and achieve rapid adaptation to different user preferences. Step S700: Verify the trained charging strategy using simulation experiments, and compare it with the constant current and constant voltage charging strategy and the multi-stage constant current fast charging strategy based on Bayesian optimization to verify the effectiveness and superiority of the proposed optimization strategy.

[0058] Furthermore, in step S100, the power battery simulation model constructed by this invention is a fractional-order electro-thermal coupling model, such as... Figure 2As shown in the figure. In this embodiment, the battery's electrical model adopts a fractional-order equivalent circuit model to describe the polarization effect and memory characteristics of the battery during charging. The electrical model includes an open-circuit voltage source, an ohmic internal resistance, and multiple fractional-order RC branches. Simultaneously, a one-dimensional lumped-parameter thermal model is constructed to describe the temperature change process caused by energy loss during charging. The thermal model describes the heat transfer relationship between the battery's interior, surface, and environment through thermal resistance and thermal capacity parameters.

[0059] Furthermore, the electrical and thermal models are coupled through heat generation terms: the voltage and current information calculated by the electrical model is used to calculate the heat generation rate, serving as input to the thermal model; the temperature results calculated by the thermal model are, in turn, used to correct temperature-related parameters in the electrical model, thereby achieving dynamic coupling between electrical and thermal behaviors. Through parameter identification and model verification, a fractional-order electro-thermal coupling model applicable to a wide temperature range is obtained.

[0060] Further, in step S200, the electrical and thermal performance of the power battery are tested under different ambient temperature conditions to obtain data for the subsequent construction of a fractional-order electro-thermal coupling model. In this embodiment, the test ambient temperature covers a wide operating temperature range of the power battery, and the collected data includes at least: sequential data of charging current, terminal voltage, and battery surface temperature changes over time.

[0061] Furthermore, such as Figure 3 As shown, step S100 includes the following test items: (1) Capacity test of the battery after full charge at room temperature and constant current discharge at 1C. (2) Perform operating condition tests on the hybrid power pulse characteristic (HPPC) at room temperature; (3) Conduct open-circuit voltage tests under low-rate conditions to establish the correspondence between open-circuit voltage and state of charge; (4) Collect voltage and current response data under dynamic charging conditions for identification and verification of fractional-order electrical model parameters; (5) Conduct charging temperature rise tests under different charging rate conditions to identify thermal model parameters.

[0062] Through the above testing process, a data set reflecting the electrical and thermal behavior of the power battery under different temperatures and charging conditions is obtained. For example... Figure 4-5As shown, in step S200, a fractional-order electro-thermal coupling model of the power battery is constructed based on the experimental data obtained in step S100. In this embodiment, the battery's electrical model adopts a fractional-order equivalent circuit model to describe the dynamic voltage changes of the power battery during charging due to polarization and memory effects. The fractional-order electrical model includes an open-circuit voltage source, an ohmic internal resistance, and multiple fractional-order RC branches, and its model parameters are related to the battery temperature. Simultaneously, a one-dimensional lumped-parameter thermal model is constructed to describe the temperature change process of the power battery during charging due to energy loss. The thermal model describes the heat transfer relationship between the battery's interior, surface, and environment through thermal resistance and thermal capacity parameters. Further, the electrical model and the thermal model are coupled through a heat generation term: the voltage and current output by the electrical model are used to calculate the heat generation rate during charging, and the heat generation rate is used as the input of the thermal model to calculate the battery temperature change; the battery temperature calculated by the thermal model is used to update the temperature-related parameters in the electrical model, thereby achieving dynamic coupling between electrical and thermal behaviors. Through parameter identification and fitting verification, a fractional-order electro-thermal coupling model applicable to a wide temperature range is obtained.

[0063] Furthermore, in step S300, based on the electro-thermal coupling model constructed in step S200, the fast charging control problem of the power battery is modeled as a multi-objective optimization problem.

[0064] In this implementation method, the optimization objectives include at least the following three aspects: 1) Shorten the charging time required for the power battery to reach the target state of charge; 2) Reduce energy loss during the charging process; 3) Suppress battery temperature rise or the rate of temperature rise to ensure thermal safety.

[0065] A comprehensive optimization objective function is constructed by weighting and combining the above-mentioned multiple optimization objectives, where the weights corresponding to each objective term represent the importance of different optimization objectives. Simultaneously, safety constraints are applied during the optimization process, including charging current constraints, upper limit constraints on battery terminal voltage, and upper limit constraints on battery temperature, to ensure that the charging process remains within a safe range.

[0066] Further, in step S400, the multi-objective optimization problem constructed in step S300 is transformed into a reinforcement learning decision problem. In this implementation method, the state of charge, terminal voltage, and battery temperature of the power battery are used as the state inputs of the reinforcement learning algorithm, and the charging current is used as the action output of the reinforcement learning algorithm. The reinforcement learning agent interacts with the electro-thermal coupling model constructed in step S200, continuously trying different charging current strategies in the simulation environment. During training, reward values ​​are calculated based on charging time, energy loss, and temperature rise performance, and penalty terms are introduced when overvoltage or overtemperature occurs to guide the reinforcement learning agent to learn a fast charging strategy that meets safety constraints. Through continuous iterative training, a charging control strategy that can achieve multi-objective optimization under the condition of meeting safety constraints is obtained.

[0067] Furthermore, in step S500, by adjusting the weights corresponding to each objective item in the multi-objective optimization, user preference-adaptive fast charging optimization control is achieved. In this implementation method, different weight combinations correspond to different user charging preferences, such as: a preference for shortening charging time; a preference for reducing temperature rise and improving safety; and a preference for reducing energy loss and improving charging efficiency. By changing the weight configuration, the reinforcement learning algorithm generates corresponding optimized charging strategies under different user preference conditions. Figure 6 (a)-(h), Figure 7 (a)-(h) and Figure 8 (a)-(c), by changing the weights of different optimization objectives, the trained agent will exhibit different fast charging optimization preferences.

[0068] Furthermore, in step S600, a transfer learning strategy is employed to achieve rapid adaptation under different user preference conditions. In this implementation method, the reinforcement learning model is first fully trained under a certain user preference condition to obtain a basic charging strategy; when user preferences change, the trained model parameters are transferred as initial parameters to the new user preference condition, and the strategy adjustment is completed through a small amount of retraining. Since the state space and action space of the old and new tasks remain consistent, and only the reward weights change, the transfer learning process can significantly reduce the time and computational cost required for retraining. Figure 9 (a)-(d) By adopting the transfer learning strategy, the fast charging strategy can be quickly transferred to specific preferences through a small amount of secondary training on the basis of pre-training, thereby meeting the needs of different users for rapid adaptation.

[0069] Further, in step S700, the trained charging strategy is simulated and verified based on the constructed electro-thermal coupling model. In this embodiment, the charging strategy is compared with the traditional constant current constant voltage charging strategy and the multi-stage constant current charging strategy based on Bayesian optimization. The different strategies are comprehensively evaluated from the perspectives of charging time, energy loss, and temperature rise control. Simulation results show that the fast charging method proposed in this invention can achieve better overall charging performance while ensuring battery safety, and has good adaptability to user preferences.

[0070] In other embodiments, the present invention also provides an electronic device, including a memory and a processor; wherein the memory stores a computer program that, when executed by the processor, enables the above-described user preference-based fast charging method for a power battery that considers temperature rise constraints.

[0071] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described experimental data-driven reinforcement learning fast charging control method for lithium-ion batteries.

[0072] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A user preference-based fast adaptive power battery charging method considering temperature rise constraints, characterized in that, include: The voltage and temperature data of the power battery under different charging currents and ambient temperatures are obtained, and the voltage and temperature data are used to construct an electro-thermal coupling model of the power battery. With the optimization goals of shortening charging time, reducing energy loss during charging, and suppressing battery temperature rise or temperature rise rate, a reinforcement learning algorithm is used to iteratively optimize different charging current strategies to obtain the basic charging strategy under the target user's preference conditions. During the iterative optimization process, the electric-thermal coupling model of the power battery is used to calculate the state transition of the charging action, and the reward value is output based on the reward function. When user preferences change, the weights of different target items in the reward function are adjusted by the user preference weight parameter, and a transfer learning strategy is adopted to transfer the basic charging strategy to different user preference conditions for secondary iterative optimization to obtain the optimal charging current strategy.

2. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 1, characterized in that, Obtaining the voltage and temperature data of the power battery under different charging currents and ambient temperatures includes: The power battery was tested for electrical and thermal performance under multiple ambient temperature conditions. The current, voltage and external surface temperature information of the battery during the charging process were collected to obtain voltage change characteristics and temperature change characteristics as the voltage and temperature data.

3. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 1, characterized in that, The power battery electro-thermal coupling model includes: A fractional-order equivalent circuit model is used to describe the dynamic electrical characteristics of the power battery during charging, where the voltage changes with the battery's state of charge and charging current. A one-dimensional lumped parameter thermal model is used to describe the temperature change process of the power battery during charging.

4. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 3, characterized in that, Constructing the electro-thermal coupling model of the power battery includes: The heat generation rate is calculated using the fractional-order equivalent circuit model. The heat generation rate is then input into the one-dimensional lumped parameter thermal model to obtain temperature results. These results are used to modify the temperature-related parameters of the fractional-order equivalent circuit model, thereby dynamically coupling electrical and thermal behaviors to generate the power battery electro-thermal coupling model.

5. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 1, characterized in that, The optimization objectives include: ； in, This indicates the cost of charging time. Indicates the cost of temperature rise. Indicates the cost of energy loss. These represent the weights of each cost.

6. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 1, characterized in that, The method also includes: Safety constraints are set for the charging process, including: charging current constraints, upper limit constraints for battery terminal voltage, upper limit constraints for battery temperature, and state of charge range constraints.

7. The user preference-based fast adaptive power battery charging method considering temperature rise constraints according to claim 1, characterized in that, The reward function includes: ； in, This indicates the cost of charging time. Indicates the cost of temperature rise. Indicates the cost of energy loss. These represent the weights of each cost. and These represent the over-temperature threshold penalty and the over-voltage threshold penalty, respectively.

8. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements each step of the charging method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the charging method according to any one of claims 1 to 7.

Images