Battery charging protection method and system based on intelligent control and storage medium

Abstract

The invention relates to the technical field of battery protection control, and discloses a battery charging protection method and system based on intelligent control and a storage medium. The method comprises the following steps: acquiring and calculating battery parameters to obtain a state value; constructing a neural network model based on the state value to output a charging parameter; generating a charging strategy according to the charging parameters; the charging strategy is input into the control circuit for execution; executing multi-stage safety protection to obtain a result; and updating neural network model optimization control parameters according to a result. According to the invention, the charging control model based on the back-propagation neural network is constructed, and the real-time battery state monitoring, the self-adaptive charging strategy generation, the multi-stage safety protection mechanism and the self-learning optimization function are combined, so that the precise control and the comprehensive protection of the battery charging process are realized, the service life of the battery is effectively prolonged, and the charging safety is improved.

Description

Technical Field

[0001] The present application relates to the field of battery protection control technology, and in particular to a battery charging protection method, system and storage medium based on intelligent control. Background Art

[0002] With the rapid development of electric vehicles, renewable energy storage systems and portable electronic devices, the performance and life of batteries as key energy storage components have become important factors restricting the further development of these technologies. The traditional battery charging method mainly adopts the constant current-constant voltage (CC-CV) charging mode. Although it is simple to operate, it lacks the ability to adaptively adjust to different battery states and environmental conditions. Although some improved charging methods such as pulse charging and multi-stage charging have improved the charging efficiency and battery life to a certain extent, they still use fixed charging parameters and preset thresholds, and cannot be dynamically optimized according to the real-time state and aging degree of the battery. In recent years, with the development of artificial intelligence technology, battery management systems based on intelligent algorithms such as fuzzy logic and neural networks have gradually been applied to battery charging control, but most systems only focus on the optimization of charging efficiency or charging time, and do not consider battery safety protection and life extension enough.

[0003] The main problems with existing battery charging technologies include: first, the lack of comprehensive monitoring and precise evaluation of battery parameters makes it impossible to accurately grasp the battery status during the charging process, increasing the risk of overcharging and over-discharging; second, charging strategies generally use static parameter settings, which cannot be adaptively adjusted according to the dynamic changing characteristics of the battery and environmental conditions, affecting charging efficiency and safety; third, most charging systems lack effective safety protection mechanisms. When abnormal conditions occur, they either rely entirely on hardware protection to cut off charging, or only have simple parameter adjustments, and cannot achieve multi-level protection; fourth, as the battery ages, its internal characteristics change, but existing charging systems rarely consider the impact of battery aging on charging parameters, making it difficult to adapt to the management needs of the battery throughout its life cycle; finally, existing systems usually lack self-learning capabilities and are unable to summarize experience from historical charging data and optimize charging strategies, making it difficult to continuously improve charging control. Summary of the invention

[0004] The present application provides a battery charging protection method, system and storage medium based on intelligent control, which is used to achieve precise control and all-round protection of the battery charging process by constructing a charging control model based on a back-propagation neural network, combining real-time battery status monitoring, adaptive charging strategy generation, multi-level safety protection mechanism and self-learning optimization function, thereby effectively extending the battery life and improving charging safety.

[0005] In the first aspect, the present application provides a battery charging protection method based on intelligent control, and the battery charging protection method based on intelligent control includes: collecting parameters of the battery voltage, current, temperature and internal resistance to obtain battery parameter data, and calculating and processing the battery parameter data to obtain a charging state value, a health state value and a power state value; constructing a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain an optimal charging current value, a maximum allowable charging current threshold, a charging termination voltage and a charging risk level; generating a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level to obtain a charging strategy; inputting the charging strategy into a power electronic control circuit for charging control to obtain a charging control result; executing a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result; performing reinforcement learning update on the back propagation neural network model according to the safety protection result and the charging cycle data to obtain optimized battery charging protection control parameters.

[0006] In a second aspect, the present application provides a battery charging protection system based on intelligent control, and the battery charging protection system based on intelligent control includes: The acquisition module is used to acquire the parameters of the voltage, current, temperature and internal resistance of the battery to obtain the battery parameter data, and to calculate and process the battery parameter data to obtain the charging state value, health state value and power state value; A construction module, used to construct a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain an optimal charging current value, a maximum allowable charging current threshold, a charging termination voltage and a charging risk level; A generating module, configured to generate a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, and obtain a charging strategy; A control module, used for inputting the charging strategy into a power electronic control circuit to perform charging control and obtain a charging control result; An execution module, configured to execute a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result; An updating module is used to perform reinforcement learning update on the back propagation neural network model according to the safety protection results and charging cycle data to obtain optimized battery charging protection control parameters.

[0007] In a third aspect, a battery charging protection device based on intelligent control is provided, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the battery charging protection device based on intelligent control executes the above-mentioned battery charging protection method based on intelligent control.

[0008] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium is run on a computer, the computer executes the above-mentioned battery charging protection method based on intelligent control.

[0009] In the technical solution provided by the present application, by collecting parameters of the voltage, current, temperature and internal resistance of the battery, and combining the charging state value, health state value and power state value to construct a back propagation neural network model, significant beneficial effects in many aspects are achieved. First, the application of the back propagation neural network model transforms the battery charging control from the traditional fixed parameter mode to the adaptive intelligent control mode. The nonlinear mapping ability of the neural network can accurately capture the complex relationship between the battery parameters and output the optimized charging control parameters, including the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, which significantly improves the accuracy and adaptability of the charging control. Secondly, the adaptive generation technology of the segmented charging curve dynamically adjusts the parameters of each stage of pre-charging, constant current charging, constant voltage charging and trickle charging according to the real-time state of the battery, taking into account the charging efficiency and battery safety, and effectively reducing the damage to the battery during the charging process. Third, the precise coordination of the power electronic control circuit and the charging strategy realizes the precise execution of the charging process through high-precision closed-loop control, ensuring the effective implementation of the charging strategy. Fourth, the implementation of the multi-level safety protection mechanism builds a comprehensive safety protection system at both the software and hardware levels, which can quickly respond to various abnormal situations, from minor parameter adjustments to emergency charging interruptions, providing gradient protection measures and significantly improving the safety of the charging process. Finally, the reinforcement learning update mechanism enables the system to have self-learning optimization capabilities. By continuously accumulating charging experience and optimizing the neural network model, the system performance continues to improve with the use time. At the same time, it can adapt to the battery aging characteristics and provide optimized charging management for the entire life cycle of the battery. The characteristics of the back propagation neural network algorithm in the present invention contribute significantly to the scheme. Its multi-layer neuron structure can learn the complex nonlinear relationship in the battery charging process. The ReLU activation function improves the expression ability of the model. The L2 regularization method prevents overfitting and ensures the generalization ability of the model. The parameter optimization based on gradient descent enables the model to be continuously adjusted to adapt to changes in battery characteristics. These algorithm features together form an adaptive, precise and safe intelligent battery charging control system, which solves the problems of fixed parameters, insufficient safety protection and poor adaptability in traditional charging methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 A schematic diagram of an embodiment of a battery charging protection method based on intelligent control in an embodiment of the present application; Figure 2 A schematic diagram of an embodiment of a battery charging protection system based on intelligent control in an embodiment of the present application; Figure 3 It is a schematic block diagram of the structure of a battery charging protection device based on intelligent control in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] Embodiments of the present application provide a battery charging protection method, system and storage medium based on intelligent control. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the battery charging protection method based on intelligent control includes: Step S101, collecting parameters of the voltage, current, temperature and internal resistance of the battery to obtain battery parameter data, and calculating and processing the battery parameter data to obtain a charging state value, a health state value and a power state value; Step S102: construct a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level; Step S103, generating a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level to obtain a charging strategy; Step S104, inputting the charging strategy into the power electronic control circuit to perform charging control and obtain a charging control result; Step S105: Execute a multi-level safety protection mechanism based on the battery parameter data, charging strategy and charging control result to obtain a safety protection result; Step S106: Perform reinforcement learning update on the back propagation neural network model according to the safety protection results and the charging cycle data to obtain optimized battery charging protection control parameters.

[0014] It is understandable that the execution subject of the present application can be a battery charging protection system based on intelligent control, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0015] Specifically, the voltage, current, temperature and internal resistance of the battery are collected through a high-precision data acquisition circuit. In this process, a 16-bit ADC converter is used to collect data at a sampling frequency of 100Hz to ensure the accuracy and real-time nature of the data. The temperature sensor uses a PT100 thermistor, and the current acquisition uses a Hall current sensor. The collected raw data is processed by a Butterworth low-pass filter with a cutoff frequency of 10Hz to filter out high-frequency noise interference and form battery parameter data. Then, the charging state value is calculated using the improved ampere-hour measurement method combined with the Kalman filter algorithm. This method integrates the charging and discharging current and combines it with the open-circuit voltage correction, and the accuracy can reach ±2%. The health state value is calculated by a multi-parameter fusion method, which comprehensively considers the battery capacity attenuation rate, the internal resistance increase rate and the charge and discharge efficiency change rate to generate a 0-100% health score. The power state value calculates the maximum allowable charging power based on the current battery temperature, charging state value and internal resistance. A back propagation neural network model is constructed based on the obtained battery parameter data, charging state value, health state value and power state value. The model adopts a four-layer structure, including an input layer, two hidden layers and an output layer. The input layer contains 10 neurons, corresponding to the key parameters in the previous step. The first hidden layer contains 16 neurons, and the second hidden layer contains 8 neurons. The ReLU activation function is used to enhance the nonlinear expression ability of the network. The output layer contains 4 neurons, which output the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level respectively. The neural network is trained by the back propagation algorithm, the loss function selects the mean square error, the optimizer uses the Adam algorithm, and the training data set consists of historical charging records. The L2 regularization method is used to prevent overfitting during training, the regularization coefficient is set to 0.0001, and an early stopping mechanism is introduced.

[0016] A segmented charging curve is generated based on the output optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level. The charging process is divided into a pre-charging stage, a constant current charging stage, a constant voltage charging stage and a trickle charging stage. In the pre-charging stage, the charging current is set to 0.1 times the maximum allowable charging current threshold. In the constant current charging stage, the optimal charging current value is used but does not exceed the maximum allowable threshold. In the constant voltage charging stage, the voltage is constant at the charging termination voltage value. In the trickle charging stage, a pulse charging method is used with a frequency of 1kHz. According to different charging risk levels, the system automatically adjusts the parameters of each stage to form a complete charging strategy. The charging strategy is input into the power electronic control circuit for charging control. The control circuit is based on a digital signal processor and combines a field programmable gate array to achieve high-speed signal processing and PWM generation. The power conversion circuit adopts a full-bridge LLC resonant conversion topology, and the main switch tube adopts a silicon carbide MOSFET. The control circuit implements closed-loop control through a PI control algorithm, and automatically switches the control mode according to the charging stage: current closed-loop control is used in the pre-charging and constant current charging stages, voltage closed-loop control is used in the constant voltage charging stage, and the power switch tube is controlled by a PWM signal in the trickle charging stage.

[0017] A multi-level safety protection mechanism is implemented based on battery parameter data, charging strategy and charging control results. The multi-level safety protection includes two levels: software protection and hardware protection. The software protection layer performs a state assessment every 100ms, including abnormal charging current detection, abnormal charging voltage detection, abnormal temperature detection and abnormal charging state value change rate detection. When the charging current exceeds the maximum allowable charging current threshold, a current abnormality alarm is triggered; when the charging voltage exceeds the charging termination voltage, a voltage abnormality alarm is triggered; when the temperature exceeds the set threshold, the system automatically adjusts the charging current or stops charging according to the temperature level. The hardware protection layer implements overcurrent protection, overvoltage protection and overtemperature protection through the power electronic control circuit, with a response time of less than 10μs. The back propagation neural network model is updated through reinforcement learning based on the safety protection results and charging cycle data. The safety protection results and charging cycle data are combined into a training sample set corresponding to the battery response characteristics under different charging conditions. After the data is normalized to eliminate the dimensionality effect, it is input into the back propagation neural network model to calculate the difference between the actual charging effect and the expected effect to obtain the model error. Based on the model error, the network weight parameters are adjusted by gradient descent to update the network parameters. Perform time series analysis on the battery capacity and internal resistance in the battery parameter data, calculate the capacity attenuation trend and internal resistance change trend, make aging compensation adjustments to the updated network parameters, and apply them to the control parameter prediction for the next charging cycle.

[0018] For example, in a specific battery charging protection application scenario, when the battery voltage is 3.2V, the current is 2A, the surface temperature is 25℃, and the internal resistance is 50mΩ, stable battery parameter data is obtained after filtering. The charging state value is calculated to be 30% by the ampere-hour measurement method combined with the Kalman filter algorithm, the health state value is 85% by the multi-parameter fusion method, and the power state value is 100W based on the current temperature and internal resistance. After these data are input into the back propagation neural network model, the optimal charging current is output as 2.5A, the maximum allowable charging current threshold is 3A, the charging termination voltage is 4.2V, and the charging risk level is low risk. Based on this, a segmented charging curve is generated: the current in the pre-charging stage is 0.3A until the battery voltage reaches 3.8V, the current in the constant current charging stage is 2.5A until the battery voltage reaches 4.1V, the voltage in the constant voltage charging stage is constant at 4.2V until the current drops to 0.25A, and then enters the trickle charging stage with 1kHz pulse charging. During the charging process, the multi-level safety protection mechanism continuously monitors and automatically reduces the charging current to 2A when the battery temperature is detected to rise to 45°C until the temperature returns to normal. After charging is completed, the data of the entire charging process is used for reinforcement learning updates of the neural network model. The optimized model can more accurately predict the parameters required for the next charge, ensuring the safety of battery charging and maximizing the lifespan.

[0019] In the embodiment of the present application, by collecting the parameters of the voltage, current, temperature and internal resistance of the battery, and combining the charging state value, health state value and power state value to build a back propagation neural network model, many significant beneficial effects are achieved. First, the application of the back propagation neural network model transforms the battery charging control from the traditional fixed parameter mode to the adaptive intelligent control mode. The nonlinear mapping ability of the neural network can accurately capture the complex relationship between the battery parameters, and output the optimized charging control parameters, including the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, which significantly improves the accuracy and adaptability of the charging control. Secondly, the adaptive generation technology of the segmented charging curve dynamically adjusts the parameters of each stage of pre-charging, constant current charging, constant voltage charging and trickle charging according to the real-time state of the battery, taking into account the charging efficiency and battery safety, and effectively reducing the damage to the battery during the charging process. Third, the precise coordination of the power electronic control circuit and the charging strategy realizes the precise execution of the charging process through high-precision closed-loop control, ensuring the effective implementation of the charging strategy. Fourth, the implementation of the multi-level safety protection mechanism builds a comprehensive safety protection system at both the software and hardware levels, which can quickly respond to various abnormal situations, from minor parameter adjustments to emergency charging interruptions, providing gradient protection measures and significantly improving the safety of the charging process. Finally, the reinforcement learning update mechanism enables the system to have self-learning optimization capabilities. By continuously accumulating charging experience and optimizing the neural network model, the system performance continues to improve with the use time. At the same time, it can adapt to the battery aging characteristics and provide optimized charging management for the entire life cycle of the battery. The characteristics of the back propagation neural network algorithm in the present invention contribute significantly to the scheme. Its multi-layer neuron structure can learn the complex nonlinear relationship in the battery charging process. The ReLU activation function improves the expression ability of the model. The L2 regularization method prevents overfitting and ensures the generalization ability of the model. The parameter optimization based on gradient descent enables the model to be continuously adjusted to adapt to changes in battery characteristics. These algorithm features together form an adaptive, precise and safe intelligent battery charging control system, which solves the problems of fixed parameters, insufficient safety protection and poor adaptability in traditional charging methods.

[0020] In a specific embodiment, the process of executing step S101 may specifically include the following steps: The voltage, current, temperature and internal resistance of the battery are acquired through a high-precision data acquisition circuit to obtain the original analog signal; The original analog signal is input into a 16-bit ADC converter for 100 Hz frequency digital processing to obtain battery parameter data; The battery parameter data is filtered out with a Butterworth low-pass filter at a cutoff frequency of 10 Hz to obtain filtered battery parameter data; The filtered battery parameter data is input into the improved ampere-hour measurement method combined with the Kalman filter algorithm to calculate the charging state and obtain the charging state value; The filtered battery parameter data is subjected to battery health analysis through multi-parameter fusion method to obtain the health status value; The maximum power is calculated based on the temperature, charging state value and internal resistance in the filtered battery parameter data to obtain the power state value.

[0021] Specifically, the voltage, current, temperature and internal resistance of the battery are acquired through a high-precision data acquisition circuit to obtain the original analog signal. In this process, voltage acquisition is usually measured using a precision differential amplifier with a measurement range of 0-5V; current acquisition uses a Hall current sensor, which can measure the current value in the range of 0-100A without contact; temperature acquisition uses a PT100 thermistor with a measurement range of -20℃ to 80℃; internal resistance acquisition is measured by a small signal AC excitation method, that is, a small signal AC current is applied to the battery, and the voltage response is measured to calculate the internal resistance value. These acquisition components are connected to the data acquisition circuit through shielded wires to reduce environmental interference. The original analog signal is input into a 16-bit ADC converter for 100Hz frequency digital processing to obtain battery parameter data. The resolution of the 16-bit ADC converter is 2^16=65536 different digital values, which makes the voltage measurement accuracy reach 0.1mV level. The sampling frequency of 100Hz is sufficient to capture the dynamic changes of battery parameters without generating too much data processing burden. The digitized data is stored in an ordered array, including sampling timestamp, voltage value, current value, temperature value and internal resistance value.

[0022] The battery parameter data is filtered with a Butterworth low-pass filter at a cutoff frequency of 10Hz to obtain the filtered battery parameter data. The Butterworth low-pass filter is a filter with a maximum flat amplitude-frequency characteristic, which can effectively filter out high-frequency noise above 10Hz and retain the actual change trend of the battery parameters. The filtered data is smoother, reducing the data fluctuation caused by measurement noise, and providing a stable data basis for subsequent parameter calculations. The filtered battery parameter data is input into the improved ampere-hour measurement method combined with the Kalman filter algorithm to calculate the charging state and obtain the charging state value. The improved ampere-hour measurement method is a method of calculating the charge change by integrating the battery charging and discharging current, and correcting the SOC in combination with the open circuit voltage. The Kalman filter algorithm is a recursive estimation algorithm that continuously optimizes the SOC estimate through a prediction-correction process to reduce the cumulative error. In actual calculations, the calculation of the charging state value follows the following formula:

[0023] in, Indicates the maximum allowed charging power (power state value), Indicates the maximum charging power based on temperature limit, Indicates the maximum charging power based on the charging state value limit. Indicates the maximum charging power based on internal resistance limitation. These three parameters are calculated in the following ways: P_{t} is determined based on the difference between the current temperature and the safe temperature range. The closer the temperature is to the upper limit, The smaller; Determined by the charging state value, the higher the charging state value, The smaller it is, the less likely it is to overcharge; Determined by the internal resistance value, the larger the internal resistance, The smaller it is, the less likely it is to overheat.

[0024] The battery health analysis is performed on the filtered battery parameter data by the multi-parameter fusion method to obtain the health status value. The multi-parameter fusion method comprehensively considers the battery capacity decay rate, internal resistance increase rate and charge and discharge efficiency change rate, and calculates the health status value by weighted average. The battery capacity decay rate is calculated by comparing the ratio of the current maximum charging capacity to the initial capacity; the internal resistance increase rate is calculated by comparing the ratio of the current internal resistance to the initial internal resistance; the charge and discharge efficiency change rate is calculated by measuring the energy loss in the charge and discharge cycle. After normalization, these parameters are assigned different weight coefficients for weighted average to obtain a health status value in the range of 0-100%. The maximum power is calculated based on the temperature, charge status value and internal resistance in the filtered battery parameter data to obtain the power status value. The power status value indicates the maximum charging power that the current battery can safely accept, which is an important constraint condition for charging control. By comprehensively considering the impact of temperature on battery safety, the restriction of charge status on overcharging, and the impact of internal resistance on heat generation, the minimum power value under the three restrictions is selected as the final power status value to ensure that the charging process is always within a safe range.

[0025] For example, when the battery temperature is 35°C, the state of charge value is 80%, and the internal resistance is 60mΩ, the maximum charging power P_{t} under the temperature limit is first calculated to be 120W (because the critical temperature of 40°C has not been reached), the maximum charging power P_{s} under the charge state value limit is 80W (because the charge state has reached 80%, and the charging power needs to be reduced), and the maximum charging power P_{i} under the internal resistance limit is 100W (because the internal resistance value is large and the heat needs to be controlled). According to the principle of taking the minimum value of the three, the power state value finally calculated is 80W, which means that the current battery charging power should not exceed 80W to ensure charging safety and extend battery life. In this way, through the precise collection and scientific calculation of battery parameters, a reliable data basis is provided for subsequent charging control decisions based on the back propagation neural network model.

[0026] In a specific embodiment, the process of executing step S102 may specifically include the following steps: The battery parameter data, charging state value, health state value and power state value are organized into 10 input neurons as input parameters, and a four-layer neural network structure including an input layer, two hidden layers and an output layer is constructed to obtain a back propagation neural network structure model; 16 neurons and a ReLU activation function are set for the first hidden layer of the back-propagation neural network structure model, 8 neurons and a ReLU activation function are set for the second hidden layer of the back-propagation neural network structure model, and 4 neurons are set for the output layer of the back-propagation neural network structure model to obtain an initial back-propagation neural network model; Input the historical charging record data into the initial back propagation neural network model, perform back propagation training through the mean square error loss function and Adam optimizer to obtain the initial training model; The initial training model was subjected to overfitting suppression using the L2 regularization method with a regularization coefficient of 0.0001 to obtain a back-propagation neural network model; The back propagation neural network model is applied to real-time battery data processing to predict the charging control parameters and obtain the optimal charging current value, maximum allowable charging current threshold, charging termination voltage and charging risk level.

[0027] Specifically, the battery parameter data, charging state value, health state value and power state value are organized into 10 input neurons as input parameters, and a four-layer neural network structure including an input layer, two hidden layers and an output layer is constructed to obtain a back-propagation neural network structure model. The 10 input neurons include battery voltage, current, surface temperature, ambient temperature, internal resistance value, charging state value, health state value, power state value, charging time and historical charging cycle number. These parameters reflect the current state and historical usage of the battery together, providing comprehensive input information for the neural network. The design of the input layer fully considers the key influencing factors in the battery charging process. Each input parameter is normalized to unify the parameters of different dimensions into the [0,1] interval to avoid some parameters dominating the network learning process due to large values. 16 neurons and ReLU activation function are set for the first hidden layer of the back-propagation neural network structure model, 8 neurons and ReLU activation function are set for the second hidden layer of the back-propagation neural network structure model, and 4 neurons are set for the output layer of the back-propagation neural network structure model to obtain the initial back-propagation neural network model. The expression of the ReLU (Rectified Linear Unit) activation function is f(x)=max(0,x). Compared with the traditional sigmoid or tanh activation function, ReLU is simpler to calculate, and there is no gradient vanishing problem when the input is positive, which helps to accelerate the network training process. The 16 neurons in the first hidden layer make the network complex enough to capture the nonlinear relationship between the input parameters, and the 8 neurons in the second hidden layer further extract high-level features. The 4 neurons in the output layer correspond to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, and directly output the key parameters required for charging control.

[0028] The historical charging record data is input into the initial back propagation neural network model, and the back propagation training is performed through the mean square error loss function and the Adam optimizer to obtain the initial training model. The historical charging record data contains a large number of battery charging processes and results under different conditions, totaling about 100,000 groups of samples, of which 80% are used for training and 20% are used for verification. The mean square error (MSE) loss function calculates the average of the sum of squares of the difference between the predicted value and the actual value, which is suitable for regression problems and can quantify the accuracy of model prediction. The Adam optimizer is an optimization algorithm with adaptive learning rate, which combines the advantages of the momentum method and RMSProp. The learning rate is initially set to 0.001, which can automatically adjust the learning rate according to the gradient change, accelerate convergence and jump out of the local optimum. The training process adopts batch processing, with 64 samples per batch, to reduce memory usage and increase training speed. During the training process, the gradient is calculated and the network weights are updated by back propagation, the prediction error is continuously reduced, and the model performance is gradually optimized. The initial training model is suppressed from overfitting by the L2 regularization method with a regularization coefficient of 0.0001 to obtain the back propagation neural network model. L2 regularization constrains the weight value by adding a penalty term of the sum of squared weights to the loss function to prevent the model from overfitting the training data. The regularization coefficient of 0.0001 controls the regularization strength and balances the fitting ability and generalization ability. At the same time, an early stopping mechanism is introduced during the training process. When the error of the validation set does not decrease for 50 consecutive iterations, the training is stopped to avoid the decrease in generalization ability caused by overtraining. The model after L2 regularization has an accuracy of more than 95% on the validation set, a sensitivity of 98% for charging risk prediction, and a specificity of 96%, indicating that the model can not only correctly identify the normal charging status, but also accurately predict potential risks.

[0029] The back propagation neural network model is applied to real-time battery data processing to predict the charging control parameters and obtain the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level. In the real-time application process, the current voltage, current, temperature and other parameters of the battery are first collected, combined with the charging state value, health state value and power state value calculated in the previous step, and input into the trained neural network model after normalization. The network forward propagation calculates the weighted sum and activation value of each layer of neurons, and finally the four neurons in the output layer give the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level. Among them, the optimal charging current value indicates the current value that can ensure the charging speed and protect the battery under the current conditions; the maximum allowable charging current threshold is the current upper limit set from a safety perspective and should not exceed this value; the charging termination voltage is the charging cut-off voltage dynamically adjusted based on the current state of the battery; the charging risk level is divided into four levels: low risk, medium risk, high risk and extremely high risk, which are used to guide the adjustment of subsequent charging strategies.

[0030] For example, when charging a group of lithium-ion batteries, the battery voltage is 3.6V, the current is 0A (not yet charged), the surface temperature is 30℃, the ambient temperature is 25℃, the internal resistance is 45mΩ, and the calculated charging state value is 40%, the health state value is 85%, the power state value is 120W, the charging time is 0 minutes, and the number of historical charging cycles is 200. These data are input into the neural network after normalization. Through the nonlinear transformation of the two hidden layers, the network finally outputs the optimal charging current value of 2.2A, the maximum allowable charging current threshold is 3.5A, the charging termination voltage is 4.15V, and the charging risk level is low risk. This indicates that the current battery is in good condition and can be charged with a current of 2.2A, but it should not exceed 3.5A, and the charging voltage should not exceed 4.15V. These parameters directly guide the generation of subsequent charging strategies to ensure that the charging process is both efficient and safe, which fully reflects the core value of the battery charging protection method based on intelligent control. As the battery status changes, such as temperature increases and charging status value increases, the neural network will adjust the output parameters in real time and dynamically optimize the charging process, effectively extending the battery life and improving charging safety.

[0031] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Compare and analyze the optimal charging current value and the maximum allowable charging current threshold to obtain a safe charging current value; Divide the pre-charging stage based on the safe charging current value and the charging risk level, and obtain the pre-charging stage parameters in response to the charging control applied when the battery charging state value is lower than the specified threshold; Based on the safe charging current value and the charging risk level, the constant current charging stage is divided, and the charging control before the battery voltage reaches a specified proportion of the charging termination voltage is applied to obtain the constant current charging stage parameters; Divide the constant voltage charging stage based on the charging termination voltage, and obtain the constant voltage charging stage parameters corresponding to the charging control applied after the battery voltage reaches the charging termination voltage; Divide the trickle charging stage based on the specified ratio of the safe charging current value, and obtain the trickle charging stage parameters corresponding to the pulse charging control applied after the charging current drops to the specified threshold; The pre-charging stage parameters, constant current charging stage parameters, constant voltage charging stage parameters and trickle charging stage parameters are combined into a charging current-time curve, a charging voltage-time curve and a PWM control signal to obtain a charging strategy.

[0032] Specifically, the optimal charging current value and the maximum allowable charging current threshold are compared and analyzed to obtain the safe charging current value. The comparison and analysis process adopts a simple and intuitive minimum value selection method, that is, the smaller value of the optimal charging current value and the maximum allowable charging current threshold is taken as the safe charging current value. The purpose of this design is to take into account both charging efficiency and safety. When the optimal charging current value output by the neural network is less than the safety threshold, the optimal value is directly adopted; when the optimal value exceeds the safety threshold, the safety threshold is used to ensure that the charging current does not cause damage to the battery. The pre-charging stage is divided based on the safe charging current value and the charging risk level, and the pre-charging stage parameters are obtained for the charging control when the battery charging state value is lower than the specified threshold. The pre-charging stage is mainly for severely discharged batteries and starts when the charging state value is lower than 10%. In this stage, the charging current is set to 0.1 times the safe charging current value. The use of small current charging can effectively avoid overheating problems caused by excessive internal resistance and reduce the impact on the battery active materials. The pre-charging stage lasts until the battery voltage reaches 90% of the rated voltage. The pre-charging stage parameters include charging current value, start conditions and termination conditions. For different charging risk levels, the pre-charging current will also be adjusted accordingly: in low-risk state, the original parameters are maintained; in medium-risk state, the pre-charging current is further reduced to 0.08 times the safe charging current value; in high-risk state, it is reduced to 0.05 times; in extremely high-risk state, charging is suspended.

[0033] The constant current charging stage is divided based on the safe charging current value and the charging risk level, and the constant current charging stage parameters are obtained for the charging control before the battery voltage reaches the specified proportion of the charging termination voltage. The constant current charging stage is the main stage of battery charging. In this stage, the charging current is constant at the safe charging current value, and the battery voltage gradually increases with the charging process. The constant current charging stage lasts until the battery voltage reaches 98% of the charging termination voltage, at which time the battery has been charged with approximately 80% of its capacity. The constant current charging stage parameters include constant charging current value, start conditions, and termination conditions. Similarly, dynamic adjustment is made according to the charging risk level: in low-risk states, the original safe charging current value is maintained; in medium-risk states, the charging current is reduced to 80% of the safe charging current value; in high-risk states, it is reduced to 50%; charging is suspended in extremely high-risk states.

[0034] The constant voltage charging stage is divided based on the charge termination voltage, and the constant voltage charging stage parameters are obtained corresponding to the charging control after the battery voltage reaches the charge termination voltage. In the constant voltage charging stage, the charging voltage is kept constant at the charge termination voltage. As the battery is charged further, the charging current gradually decreases. The main purpose of this stage is to safely charge the battery to a state close to full charge to avoid damage to the battery caused by overvoltage. The constant voltage charging stage lasts until the charging current drops to 0.1 times the safe charging current value, at which time the battery capacity has reached about 95%. The constant voltage charging stage parameters include constant charging voltage value, start conditions and termination conditions. The processing under different charging risk levels is: in low risk state, the charging termination voltage given by the neural network is used; in medium risk state, the termination voltage is reduced by 0.05V; in high risk state, it is reduced by 0.1V; in extremely high risk state, charging is suspended.

[0035] The trickle charging stage is divided based on the specified ratio of the safe charging current value, and the parameters of the trickle charging stage are obtained for the pulse charging control after the charging current drops to the specified threshold. The trickle charging stage adopts a pulse charging method, the pulse width modulation frequency is set to 1kHz, and the duty cycle is dynamically adjusted according to the battery temperature, ranging from 10% to 30%. This method can avoid the damage to the battery caused by long-term small current charging while ensuring full charging. The duration of the trickle charging stage is determined according to the battery health status value. The lower the health status value, the shorter the trickle charging time, so as to reduce further damage to the aged battery. The parameters of the trickle charging stage include pulse frequency, duty cycle, duration and termination conditions. Under different charging risk levels, the trickle charging process is as follows: normal execution under low risk conditions; shortening the duration to 80% of normal under medium risk conditions; shortening to 50% under high risk conditions; directly skipping the trickle charging stage under extremely high risk conditions.

[0036] The pre-charging stage parameters, constant current charging stage parameters, constant voltage charging stage parameters and trickle charging stage parameters are combined into charging current-time curve, charging voltage-time curve and PWM control signal to obtain the charging strategy. This process connects the discrete parameter points of each stage into a continuous curve through the time series interpolation algorithm to ensure a smooth transition of the charging process. The charging current-time curve describes the charging current at each time point, the charging voltage-time curve describes the charging voltage level at each time point, and the PWM control signal is mainly used for pulse control in the trickle charging stage. These three sets of data together constitute a complete charging strategy, providing detailed control instructions for the subsequent power electronic control circuit.

[0037] For example, for a group of lithium-ion batteries, the charging control is performed. Assume that the optimal charging current value output by the back propagation neural network model is 2.5A, the maximum allowable charging current threshold is 2.2A, the charging termination voltage is 4.2V, and the charging risk level is medium risk. First, the safe charging current value is determined to be 2.2A (the minimum value) through comparative analysis. Since the battery charging state value is 5%, which is lower than the pre-charging threshold of 10%, the pre-charging stage is entered, and the charging current is set to 0.08 times the safe charging current value (because of medium risk), that is, 0.176A. Pre-charging continues until the battery voltage rises to 3.78V (90% of the rated voltage of 4.2V). Then it enters the constant current charging stage, and the charging current is set to 80% of the safe charging current value (because of medium risk), that is, 1.76A, and continues until the battery voltage reaches 4.116V (98% of the termination voltage of 4.2V). Then it enters the constant voltage charging stage, keeping the charging voltage constant at 4.15V (termination voltage minus 0.05V, because of medium risk), and continues until the charging current drops to 0.176A (0.1 times of 1.76A). Finally, it enters the trickle charging stage, using pulse charging with a frequency of 1kHz and a duty cycle of 20%, and the duration is 80% of the normal value (because of medium risk). The entire charging process is output in the form of charging current-time curve, charging voltage-time curve and PWM control signal, forming a complete charging strategy, guiding the power electronic control circuit to accurately control the charging process, and ensuring that the battery is charged efficiently under safe conditions.

[0038] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Input the charging current-time curve, charging voltage-time curve and PWM control signal in the charging strategy into the power electronic control circuit with the digital signal processor as the core to obtain the control instruction data; The control instruction data is processed at high speed and PWM is generated through a field programmable gate array to obtain a power switch control signal; Based on the power switch control signal, the full-bridge LLC resonant conversion topology using silicon carbide MOSFET is driven to obtain accurate charging current and charging voltage output; The charging current is closed-loop controlled by the PI algorithm to obtain a current control loop; The charging voltage is closed-loop controlled by the PI algorithm to obtain a voltage control loop; The current control loop and the voltage control loop are automatically switched according to the charging stage. The current control loop is used in the pre-charging and constant current stages, the voltage control loop is used in the constant voltage stage, and PWM direct control is used in the trickle current stage to obtain the charging control result.

[0039] Specifically, the charging current-time curve, charging voltage-time curve and PWM control signal in the charging strategy are input into the power electronic control circuit with the digital signal processor as the core to obtain the control instruction data. The digital signal processor (DSP) is a microprocessor specially used for digital signal processing, which has high-speed computing capability and dedicated instruction set, and is suitable for real-time control applications. In the battery charging control, the DSP adopts the TMS320F28335 chip with a main frequency of 150MHz and 12-bit ADC conversion accuracy, which can quickly process the input charging strategy curve. The DSP first reads the charging current-time curve, charging voltage-time curve and PWM control signal data stored in the memory, and searches for the corresponding target charging current, charging voltage and PWM parameters according to the current time point to form the control instruction data. The control instruction data contains the precise current value or voltage value that needs to be output at present, as well as the PWM frequency and duty cycle information used in the trickle charging stage. The control instruction data is subjected to high-speed signal processing and PWM generation through the field programmable gate array to obtain the power switch control signal. Field Programmable Gate Array (FPGA) is an integrated circuit that can program the internal logic structure according to the needs, with high parallel processing capability and precise timing control capability. In the charging control system, FPGA uses Xilinx Spartan-6 series with a clock frequency of 100MHz, which is responsible for receiving the control instruction data issued by DSP and performing high-speed signal processing. The precise PWM generation module is implemented in FPGA, which can generate PWM signals with a frequency range from several Hz to several MHz and a duty cycle resolution of 0.1%. For the pre-charging and constant current charging stages, FPGA generates corresponding PWM signals according to the output of the current control loop; for the constant voltage charging stage, PWM signals are generated according to the output of the voltage control loop; for the trickle charging stage, pulse signals are directly generated according to the PWM parameters provided by the charging strategy. These PWM signals are collectively referred to as power switch control signals, which directly control the on and off of the switch tube in the subsequent power conversion circuit. Based on the power switch control signal, the full-bridge LLC resonant conversion topology using silicon carbide MOSFET is driven to obtain accurate charging current and charging voltage output. The full-bridge LLC resonant conversion topology is a high-efficiency, low-noise power conversion circuit that includes a full-bridge switching circuit, an LLC resonant network, and an output rectifier filter circuit. Silicon carbide (SiC) MOSFET is a new type of wide-bandgap semiconductor power device that has lower on-resistance, higher switching frequency, and better temperature characteristics than traditional silicon-based MOSFETs. In the charging system, the power switch control signal drives the silicon carbide MOSFET through an optoelectronic isolation drive circuit to control its on and off time, thereby adjusting the output charging current and voltage. The LLC resonant network consists of a resonant inductor, a magnetizing inductor, and a resonant capacitor, which can achieve soft switching, reduce switching losses, and improve conversion efficiency.The switching frequency can be adjusted within the range of 100kHz-500kHz, and the output power can be controlled by changing the switching frequency or PWM duty cycle. The output end uses Schottky diode rectification, low ESR electrolytic capacitors and ceramic capacitors in parallel for filtering, converting the high-frequency square wave voltage into a stable DC voltage and current for battery charging.

[0040] The charging current is closed-loop controlled by the PI algorithm to obtain the current control loop. PI (proportional-integral) control is a classic feedback control algorithm that accurately controls the system output by adjusting the proportional coefficient and integral coefficient. In the current control loop, the actual output current is first measured by the Hall current sensor, which is compared with the target current value in the control instruction data to calculate the current error. The PI controller calculates the control amount based on the current error. The proportional term directly responds to the current error, and the integral term accumulates historical errors and eliminates static errors. The proportional coefficient of the current loop control is set to 0.05, the integral coefficient is set to 0.01, and the control bandwidth is 10kHz, which can respond quickly to current changes and remain stable. After the PI controller output is limited, it is sent to the FPGA to be converted into a PWM signal, and then the on-time of the power switch is adjusted to form a closed-loop control to ensure that the actual charging current is consistent with the target current. The charging voltage is closed-loop controlled by the PI algorithm to obtain the voltage control loop. The principle of the voltage control loop is similar to that of the current control loop, except that the control object changes from current to voltage. First, the actual output voltage is measured through a high-precision voltage divider sampling circuit, and it is compared with the target voltage value in the control instruction data to calculate the voltage error. The voltage PI controller calculates the control amount based on the voltage error. The proportional coefficient of the voltage loop is set to 0.1, the integral coefficient is set to 0.02, and the control bandwidth is 1kHz, which is lower than the current loop bandwidth. This is because voltage changes are usually slower than current changes. After the PI controller output is limited, it is also sent to the FPGA to be converted into a PWM signal, adjust the conduction time of the power switch, and form a voltage closed-loop control to ensure that the actual charging voltage is consistent with the target voltage.

[0041] The current control loop and the voltage control loop are automatically switched according to the charging stage. The current control loop is used in the pre-charging and constant current stages, the voltage control loop is used in the constant voltage stage, and the PWM direct control is used in the trickle stage to obtain the charging control result. Different charging stages require different control methods. The DSP automatically determines the current charging stage and switches the corresponding control loop according to the charging strategy and the real-time monitored battery status. In the pre-charging and constant current charging stages, the battery voltage has not yet reached the charging termination voltage. The main goal is to maintain constant current charging. Therefore, the current control loop is used to control the actual charging current at the target value. When the battery voltage rises to close to the charging termination voltage, it automatically switches to the constant voltage charging stage. At this time, the voltage control loop is used to keep the charging voltage constant at the charging termination voltage, and the charging current naturally decreases as the internal electrochemical state of the battery changes. When the charging current drops to the specified threshold, it enters the trickle charging stage. At this time, closed-loop control is no longer used, but the PWM control signal defined in the charging strategy is directly used to trickle charge the battery in a pulse manner until the charging is completed. The control results of the entire process include the actual output charging current, charging voltage and battery status data, which are recorded in real time and fed back to the safety protection mechanism to monitor the safety of the charging process and adjust the control strategy.

[0042] For example, when charging a group of large-capacity lithium batteries, the charging strategy is first sent to the DSP. The DSP reads that the current is the pre-charging stage and the target current is 0.5A. This data is converted into a control instruction and sent to the FPGA. The FPGA generates the corresponding PWM signal, and the duty cycle is initially set to 10%, driving the four silicon carbide MOSFET switches in the full-bridge LLC resonant circuit. The actual output current measured by the Hall current sensor is 0.48A, which has an error of 0.02A from the target value of 0.5A. The current PI controller calculates the control amount and increases the PWM duty cycle to 10.5% to make the actual current reach the target value of 0.5A. As the battery voltage rises to 3.7V, the system determines that it has entered the constant current charging stage, and the target current is increased to 2A. The control loop still uses current control. When the battery voltage reaches 4.1V, the system switches to the constant voltage charging stage, and turns on the voltage control loop to keep the charging voltage constant at 4.2V, while monitoring the gradual decrease of the charging current. When the charging current drops to 0.2A, the system enters the trickle charging stage and uses 1kHz, 20% duty cycle PWM direct control to complete the final charging process. The entire charging process is precisely controlled by the power electronic control circuit to ensure the safety and efficiency of charging.

[0043] In a specific embodiment, the process of executing step S105 may specifically include the following steps: Real-time monitoring of the charging current, charging voltage, temperature and charging status values ​​in the battery parameter data, performing status evaluation once every specified time to obtain abnormal status detection results; Compare and judge the charging current with the maximum allowable charging current threshold, and trigger a current abnormality alarm when the charging current is greater than the maximum allowable charging current threshold, thereby obtaining a current abnormality detection result; Compare and judge the charging voltage with the charging termination voltage, and when the charging voltage is greater than a specified ratio of the charging termination voltage, trigger a voltage abnormality alarm to obtain a voltage abnormality detection result; Perform multi-level threshold judgment on the temperature, reduce the charging current when the temperature exceeds the first threshold, further reduce the charging current when the temperature exceeds the second threshold, and immediately stop charging when the temperature exceeds the third threshold to obtain the temperature abnormality protection result; Combine the current anomaly detection results, voltage anomaly detection results and temperature anomaly protection results to build a multi-level protection mechanism of software protection layer and hardware protection layer to obtain a fault handling solution; The charging control result is adjusted based on the fault handling plan. When a recoverable fault is detected, the charging parameters are adjusted to continue charging. When an unrecoverable fault is detected, charging is terminated and an alarm signal is issued to obtain a safety protection result.

[0044] Specifically, the charging current, charging voltage, temperature and charging state value in the battery parameter data are monitored in real time, and a state evaluation is performed every specified time to obtain an abnormal state detection result. During the real-time monitoring process, the sampling frequency is set to 10Hz, that is, the battery parameter data is collected every 100 milliseconds. After the data collection is completed, the latest collected data is stored in the circular buffer, and the historical data of the last 30 seconds is retained for trend analysis. The state evaluation adopts a combination of threshold judgment method and change rate detection method. The threshold judgment is used to detect whether the parameter exceeds the safety range, and the change rate detection is used to identify the abnormal change trend of the parameter. After each state evaluation, an abnormal state detection result is generated, which includes the state flag bit and abnormal level of each parameter. The flag bit 0 indicates normal, 1 indicates warning, 2 indicates abnormal, and 3 indicates severe abnormal. The charging current is compared with the maximum allowable charging current threshold. When the charging current is greater than the maximum allowable charging current threshold, the current abnormal alarm is triggered to obtain the current abnormal detection result. The abnormal charging current detection has a three-level judgment mechanism: when the actual charging current exceeds 105% of the maximum allowable charging current threshold but does not exceed 110%, a first-level warning is triggered, and the warning event is recorded but charging continues; when it exceeds 110% but does not exceed 120%, a second-level warning is triggered, and the charging current is automatically reduced to 95% of the maximum allowable value; when it exceeds 120%, a third-level warning is triggered, and charging is immediately interrupted and an alarm is sounded. The abnormal current detection result includes the judgment result flag and the abnormal level, and also records the duration of the current exceeding the standard and the maximum exceeding range for subsequent fault diagnosis.

[0045] Compare and judge the charging voltage with the charging termination voltage. When the charging voltage is greater than the specified ratio of the charging termination voltage, the voltage abnormality alarm is triggered to obtain the voltage abnormality detection result. The charging voltage abnormality detection also adopts a three-level judgment mechanism: when the actual charging voltage exceeds 102% of the charging termination voltage but does not exceed 105%, a first-level warning is triggered, and the warning event is recorded but charging continues; when it exceeds 105% but does not exceed 110%, a second-level warning is triggered, and the constant voltage mode is immediately switched to control the voltage at the charging termination voltage value; when it exceeds 110%, a third-level warning is triggered, charging is immediately interrupted and an alarm is issued. The voltage abnormality detection result includes the judgment result flag and abnormality level, as well as the duration of the voltage exceeding the standard and the maximum exceeding range.

[0046] The temperature data is judged by multi-level thresholds. When the temperature exceeds the first threshold, the charging current is reduced. When the temperature exceeds the second threshold, the charging current is further reduced. When the temperature exceeds the third threshold, charging is stopped immediately to obtain the temperature abnormality protection result. The temperature multi-level threshold judgment mechanism sets four temperature threshold points: T1=40℃, T2=45℃, T3=50℃, and T4=55℃, which correspond to different protection measures. When the battery temperature exceeds T1 but is lower than T2, the charging current is reduced to 80% of the safe charging current value; when the temperature exceeds T2 but is lower than T3, the charging current is further reduced to 50% of the safe charging current value; when the temperature exceeds T3 but is lower than T4, the charging current is reduced to 20% of the safe charging current value; when the temperature exceeds T4, charging is stopped immediately and an alarm is triggered. The temperature abnormality protection result includes the temperature judgment result flag, the abnormality level, and the recommended charging current adjustment value.

[0047] The results of current anomaly detection, voltage anomaly detection and temperature anomaly protection are combined and processed to construct a multi-level protection mechanism of software protection layer and hardware protection layer, and a fault handling solution is obtained. The combined processing adopts the method of priority sorting and logical merging. First, the detection results are sorted from high to low according to the abnormality level. When multiple detections are abnormal at the same time, the highest level of processing measures are adopted. The software protection layer is mainly based on the safety monitoring program running in the DSP, with a response time of milliseconds, and is responsible for handling most abnormal situations; the hardware protection layer is based on the hardware comparator and protection circuit in the power electronic control circuit, with a response time of microseconds, as the last guarantee of the software protection layer. The combination of the two forms a complete multi-level protection mechanism to ensure that abnormal situations can be responded to in time under any circumstances. The fault handling solution is divided into two categories according to the nature of the abnormality: recoverable faults and unrecoverable faults. The former can restore normal charging by adjusting parameters, while the latter requires interrupting charging and waiting for manual intervention. The charging control result is adjusted based on the fault handling solution. When a recoverable fault is detected, the charging parameters are adjusted to continue charging. When an unrecoverable fault is detected, charging is terminated and an alarm signal is issued to obtain a safety protection result. For recoverable faults, such as slight temperature rise or current fluctuation, the charging process can be continued but kept within a safe range by adjusting the charging current, switching the charging mode or extending the charging time. The specific adjustment strategy is guided by the recommended parameters in the fault handling solution, such as reducing the charging current to a certain percentage of the original value, switching to the constant voltage stage in advance or extending the trickle charging time. For unrecoverable faults, such as severe overtemperature, continuous overcurrent or internal short circuit of the battery, the power output circuit is immediately shut down, the charging current is cut off, and an alarm signal is issued through an audible and visual alarm. At the same time, detailed fault information is recorded, including the fault type, occurrence time, related parameter values, etc., for subsequent analysis. The safety protection result includes the processed charging parameter adjustment value, protection status flag and detailed fault record.

[0048] For example, when charging a set of on-board power batteries, the battery parameter data is first continuously collected through high-precision sensors. Assume that during the charging process, the battery temperature is monitored to gradually rise from the initial 25°C to 42°C, exceeding the first temperature threshold T1 (40°C). The temperature anomaly protection mechanism immediately determines that the current temperature anomaly belongs to the first level, and generates a temperature anomaly protection result, including the temperature anomaly flag 1 and the recommended charging current adjustment coefficient 0.8. At the same time, the current and voltage detection are normal, and the abnormal flags are all 0. The multi-level protection mechanism combines these three test results. Since only the temperature has a first-level abnormality, the final fault handling solution is "minor temperature abnormality, a recoverable fault, and the charging current needs to be reduced to 80% of the original value". Based on this solution, the battery charging protection system adjusts the original charging current of 2.5A to 2.0A (2.5A×0.8) and continues to monitor the temperature change trend. After the adjustment, the battery temperature rise slowed down and stabilized at 43°C, without triggering a higher level of temperature threshold. During the whole process, charging can continue, and the temperature is controlled by actively reducing the charging current, ensuring the safety of the charging process, reflecting the adaptive protection capability of the battery charging protection method based on intelligent control.

[0049] In a specific embodiment, the process of executing step S106 may specifically include the following steps: The safety protection results and charging cycle data are combined into a training sample set, corresponding to the battery response characteristics under different charging conditions, to obtain the back propagation neural network model update data; Normalize the update data of the back propagation neural network model to eliminate the dimension effect and obtain standardized training data; The standardized training data is input into the back propagation neural network model, and the loss function value is calculated according to the difference between the actual charging effect and the expected charging effect to obtain the model error; Based on the model error, the weight parameters of the back propagation neural network model are adjusted by gradient descent to obtain updated network parameters; Perform time series analysis on the battery capacity and internal resistance in the battery parameter data, calculate the capacity attenuation trend and internal resistance change trend, and obtain the battery aging characteristic data; Based on the battery aging characteristic data, the updated network parameters are compensated and adjusted, and applied to the prediction of the control parameters of the next charging cycle, including the optimization calculation of the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, to obtain the optimized battery charging protection control parameters.

[0050] Specifically, the safety protection results and charging cycle data are combined into a training sample set, corresponding to the battery response characteristics under different charging conditions, to obtain the back propagation neural network model update data. The safety protection results include the processing records of various abnormal situations during the charging process, parameter adjustment records and the final charging effect evaluation; the charging cycle data includes a complete charging process record, such as the time series data of charging current, voltage, temperature, charging time, charging efficiency and other information. These two types of data are matched and integrated according to the timestamp to form a structured training sample set. Each group of samples contains input features (initial battery state, charging conditions, control parameters) and output labels (actual charging effect, safety risk rating). The sample set is divided into multiple batches, each of which contains samples from different battery states and charging conditions to ensure that the model can learn the optimal control strategy under various conditions. The back propagation neural network model update data is normalized to eliminate the dimension effect and obtain standardized training data. Normalization is a standard preprocessing step in machine learning. The purpose is to unify features of different dimensions into the same numerical range to avoid some features from dominating the training process due to their large values. The commonly used normalization method is minimum-maximum normalization, which maps the feature values ​​to the interval [0,1]. For continuous numerical features such as battery voltage, current, and temperature, the historical maximum and minimum values ​​of each feature are first determined, and then the original values ​​are mapped to the normalized interval through linear transformation. For example, the battery voltage usually varies in the range of 2.5V-4.2V. After normalization, 2.5V is mapped to 0, 4.2V is mapped to 1, and the intermediate values ​​are mapped proportionally. For discrete features, such as charging risk level, One-Hot Encoding is used to convert them into binary feature vectors. After the normalization process is completed, all features are in the same numerical order of magnitude, which is conducive to the efficient training of neural networks.

[0051] The standardized training data is input into the back propagation neural network model, and the loss function value is calculated according to the difference between the actual charging effect and the expected charging effect to obtain the model error. In the forward propagation stage, the back propagation neural network model calculates the weighted input and activation value of each layer in turn. For the 10 neurons in the input layer, the standardized input features are directly received; for the two hidden layers, each neuron first calculates the weighted sum of the input, and then obtains the output through the ReLU activation function; the 4 neurons in the output layer also calculate the weighted sum, but do not apply the activation function, and directly output the predicted value. These predicted values ​​are compared with the actual label values ​​(actual optimal charging parameters) to calculate the loss function value. The loss function uses the mean square error (MSE), which is the average of the sum of the squares of the differences between the predicted values ​​and the actual values. The model error includes not only the overall loss value, but also the individual error of each output neuron, which is used to guide subsequent parameter updates.

[0052] Based on the model error, the weight parameters of the back-propagation neural network model are adjusted by gradient descent to obtain the updated network parameters. Gradient descent is a commonly used parameter optimization algorithm in machine learning. By calculating the partial derivatives (gradients) of the loss function with respect to each parameter, the parameters are updated in the opposite direction of the gradient to reduce the loss value. In the back-propagation process, starting from the output layer, the gradients of the parameters of each layer are calculated backward layer by layer. First, the partial derivatives of the output layer error with respect to the weighted input of the output layer are calculated, and then the gradient of the output layer weights is calculated according to the chain rule. Then the error term of the hidden layer is calculated, and the gradient of the hidden layer weights is calculated. Finally, the Adam optimization algorithm is used to update all parameters. The Adam algorithm combines the advantages of the momentum method and RMSprop, and can adaptively adjust the learning rate, accelerate convergence and avoid local optimality. After each parameter update, the model performance is evaluated using the validation set. If the performance deteriorates, the learning rate is reduced or the training is stopped early to prevent overfitting.

[0053] The battery capacity and internal resistance in the battery parameter data are analyzed in time series, and the capacity decay trend and internal resistance change trend are calculated to obtain the battery aging characteristic data. Time series analysis is a method to study the law of data change over time, which plays an important role in battery aging analysis. First, the battery capacity and internal resistance data of each charging cycle are extracted from the historical charging records to form two time series. For the capacity time series, the capacity change rate between adjacent charging cycles is calculated, and the capacity decay curve is fitted, which usually presents an approximate exponential decay characteristic. For the internal resistance time series, the change rate is also calculated and the internal resistance growth curve is fitted, which usually presents an approximate linear or exponential growth characteristic. Combining the two curves, a battery aging characteristic model is established, which can predict the changing trend of battery capacity and internal resistance in future charging cycles. The aging characteristic data includes key indicators such as the current capacity retention rate, internal resistance growth rate, and the expected number of remaining cycles.

[0054] Based on the battery aging characteristic data, the updated network parameters are compensated and adjusted, and applied to the control parameter prediction of the next charging cycle, including the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level optimization calculation, to obtain the optimized battery charging protection control parameters. Compensation adjustment is the process of adaptively correcting the neural network model according to the battery aging characteristics. Specifically, for aged batteries, a more conservative charging strategy is required. First, the maximum allowable charging current threshold is linearly reduced according to the capacity decay rate. For every 10% decrease in capacity, the current threshold is correspondingly reduced by 5%. Secondly, the optimal charging current value is adjusted according to the internal resistance growth rate. For every 20% increase in internal resistance, the optimal charging current value is reduced by 10% to reduce the heat generated during the charging process. Thirdly, the charging termination voltage is fine-tuned according to the comprehensive aging conditions. For batteries with severe aging, the termination voltage is appropriately reduced, such as the termination voltage is reduced by 0.5% after every 100 cycles. Finally, the judgment criteria for the charging risk level are adjusted according to the aging conditions, so that aging batteries are more likely to trigger higher-level risk warnings. These compensation adjustments act directly on the output layer of the neural network model, and the output results can be dynamically adjusted without changing the network structure.

[0055] For example, when a set of lithium batteries for electric bicycles are intelligently charged, a complete charging cycle is first completed, and the battery parameter data and safety protection results of the whole process are recorded. The data shows that the battery has a slight temperature anomaly during the charging process. The safety protection mechanism reduces the charging current from the initial 2A to 1.6A, successfully controls the temperature, and the final charging capacity is 10Ah. These data are compared with the expected charging effect (target temperature range, charging time, charging capacity, etc.), and it is found that the actual temperature control effect is good, but the charging time is slightly longer than expected. This information is organized into training samples, which are normalized and then input into the neural network model. The model calculates the mean square error between the predicted value and the actual value as 0.015, and calculates the gradient of the parameters of each layer through the back propagation algorithm. The network parameters are updated using the Adam optimizer with a learning rate of 0.001, and convergence is achieved after 20 iterations. At the same time, by analyzing the historical charging data, it was found that the battery had completed 200 charge and discharge cycles, and the capacity dropped from the initial 12Ah to 10Ah, with a decay rate of 16.7%; the internal resistance increased from the initial 25mΩ to 35mΩ, with a growth rate of 40%. Based on these aging characteristic data, the output of the neural network model was compensated and adjusted: the maximum allowable charging current threshold was reduced from 2.5A to 2.3A (reduced by 8.3%), the optimal charging current value was reduced from 2A to 1.8A (reduced by 10%), and the charging termination voltage was fine-tuned from 4.2V to 4.18V. These optimized battery charging protection control parameters are applied to the next charging cycle, which not only takes into account the actual response characteristics of the battery, but also adapts to the aging condition of the battery, realizing precise control and safety protection of the charging process.

[0056] The above describes the battery charging protection method based on intelligent control in the embodiment of the present application. The following describes the battery charging protection system based on intelligent control in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the battery charging protection system based on intelligent control includes: The acquisition module is used to acquire the parameters of the voltage, current, temperature and internal resistance of the battery to obtain the battery parameter data, and to calculate and process the battery parameter data to obtain the charging state value, health state value and power state value; A construction module, used to construct a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain an optimal charging current value, a maximum allowable charging current threshold, a charging termination voltage and a charging risk level; A generating module, configured to generate a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, and obtain a charging strategy; A control module, used for inputting the charging strategy into a power electronic control circuit to perform charging control and obtain a charging control result; An execution module, configured to execute a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result; An updating module is used to perform reinforcement learning update on the back propagation neural network model according to the safety protection results and charging cycle data to obtain optimized battery charging protection control parameters.

[0057] Through the cooperation of the above components, the battery voltage, current, temperature and internal resistance are collected, and the back propagation neural network model is constructed in combination with the charging state value, health state value and power state value, achieving significant beneficial effects in many aspects. First, the application of the back propagation neural network model enables the battery charging control to be transformed from the traditional fixed parameter mode to the adaptive intelligent control mode. The nonlinear mapping ability of the neural network can accurately capture the complex relationship between the battery parameters and output the optimized charging control parameters, including the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, which significantly improves the accuracy and adaptability of the charging control. Secondly, the adaptive generation technology of the segmented charging curve dynamically adjusts the parameters of the pre-charging, constant current charging, constant voltage charging and trickle charging stages according to the real-time status of the battery, taking into account the charging efficiency and battery safety, and effectively reducing the damage to the battery during the charging process. Third, the precise coordination of the power electronic control circuit and the charging strategy realizes the precise execution of the charging process through high-precision closed-loop control, ensuring the effective implementation of the charging strategy. Fourth, the implementation of the multi-level safety protection mechanism builds a comprehensive safety protection system at both the software and hardware levels, which can quickly respond to various abnormal situations, from minor parameter adjustments to emergency charging interruptions, providing gradient protection measures and significantly improving the safety of the charging process. Finally, the reinforcement learning update mechanism enables the system to have self-learning optimization capabilities. By continuously accumulating charging experience and optimizing the neural network model, the system performance continues to improve with the use time. At the same time, it can adapt to the battery aging characteristics and provide optimized charging management for the entire life cycle of the battery. The characteristics of the back propagation neural network algorithm in the present invention contribute significantly to the scheme. Its multi-layer neuron structure can learn the complex nonlinear relationship in the battery charging process. The ReLU activation function improves the expression ability of the model. The L2 regularization method prevents overfitting and ensures the generalization ability of the model. The parameter optimization based on gradient descent enables the model to be continuously adjusted to adapt to changes in battery characteristics. These algorithm features together form an adaptive, precise and safe intelligent battery charging control system, which solves the problems of fixed parameters, insufficient safety protection and poor adaptability in traditional charging methods.

[0058] above Figure 2 The battery charging protection system based on intelligent control in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The battery charging protection device based on intelligent control in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0059] Figure 3It is a structural schematic diagram of a battery charging protection device based on intelligent control provided by an embodiment of the present invention. The battery charging protection device 300 based on intelligent control may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the battery charging protection device 300 based on intelligent control. Furthermore, the processor 310 can be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the battery charging protection device 300 based on intelligent control to implement the steps of the above-mentioned battery charging protection method based on intelligent control.

[0060] The battery charging protection device 300 based on intelligent control may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It can be understood by those skilled in the art that Figure 3 The structure of the battery charging protection device based on intelligent control shown does not constitute a limitation on the battery charging protection device based on intelligent control provided by the present invention, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0061] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the battery charging protection method based on intelligent control.

[0062] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0063] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or partly contributed to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a battery charging protection device based on intelligent control (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0064] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A battery charging protection method based on intelligent control, characterized in that: The method comprises: Collecting parameters of the voltage, current, temperature and internal resistance of the battery to obtain battery parameter data, and calculating and processing the battery parameter data to obtain a charging state value, a health state value and a power state value; Constructing a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain an optimal charging current value, a maximum allowable charging current threshold, a charging termination voltage and a charging risk level; Generate a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level to obtain a charging strategy; Inputting the charging strategy into a power electronic control circuit to perform charging control and obtain a charging control result; Execute a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result; The back propagation neural network model is updated through reinforcement learning according to the safety protection result and the charging cycle data to obtain optimized battery charging protection control parameters.

2. The battery charging protection method based on intelligent control according to claim 1 is characterized in that: The voltage, current, temperature and internal resistance of the battery are collected to obtain battery parameter data; The battery parameter data is calculated and processed to obtain a charging state value, a health state value, and a power state value, including: Acquiring signals of the voltage, current, temperature and internal resistance of the battery through a high-precision data acquisition circuit to obtain original analog signals; Inputting the original analog signal into a 16-bit ADC converter for 100 Hz frequency digital processing to obtain the battery parameter data; The battery parameter data is subjected to noise filtering with a cutoff frequency of 10 Hz through a Butterworth low-pass filter to obtain filtered battery parameter data; Inputting the filtered battery parameter data into an improved ampere-hour measurement method combined with a Kalman filter algorithm to calculate the charging state, thereby obtaining the charging state value; Performing battery health analysis on the filtered battery parameter data by a multi-parameter fusion method to obtain the health status value; The maximum power is calculated based on the temperature, the charging state value and the internal resistance in the filtered battery parameter data to obtain the power state value.

3. The battery charging protection method based on intelligent control according to claim 1 is characterized in that: The back propagation neural network model is constructed based on the battery parameter data, the charging state value, the health state value and the power state value to obtain the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, including: The battery parameter data, the charging state value, the health state value and the power state value are organized as input parameters into 10 input neurons, and a four-layer neural network structure including an input layer, two hidden layers and an output layer is constructed to obtain a back propagation neural network structure model; The first hidden layer of the back-propagation neural network structure model is provided with 16 neurons and a ReLU activation function, the second hidden layer of the back-propagation neural network structure model is provided with 8 neurons and a ReLU activation function, and the output layer of the back-propagation neural network structure model is provided with 4 neurons to obtain an initial back-propagation neural network model; Inputting the historical charging record data into the initial back propagation neural network model, performing back propagation training through a mean square error loss function and an Adam optimizer to obtain an initial training model; The initial training model is subjected to overfitting suppression by an L2 regularization method with a regularization coefficient of 0.0001 to obtain the back propagation neural network model; The back propagation neural network model is applied to real-time battery data processing to predict charging control parameters to obtain the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level.

4. The battery charging protection method based on intelligent control according to claim 1 is characterized in that: The step of generating a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level to obtain a charging strategy includes: Comparing and analyzing the optimal charging current value and the maximum allowable charging current threshold value to obtain a safe charging current value; Dividing the pre-charging stage based on the safe charging current value and the charging risk level, and obtaining pre-charging stage parameters corresponding to the charging control applied when the battery charging state value is lower than a specified threshold; Divide the constant current charging stage based on the safe charging current value and the charging risk level, and obtain constant current charging stage parameters in response to the charging control before the battery voltage reaches a specified proportion of the charging termination voltage; Divide the constant voltage charging stage based on the charging termination voltage, and obtain constant voltage charging stage parameters in response to the charging control applied after the battery voltage reaches the charging termination voltage; Divide the trickle charging stage based on the specified ratio of the safe charging current value, and obtain the trickle charging stage parameters in response to the pulse charging control applied after the charging current drops to the specified threshold; The pre-charging stage parameters, the constant current charging stage parameters, the constant voltage charging stage parameters and the trickle charging stage parameters are combined into a charging current-time curve, a charging voltage-time curve and a PWM control signal to obtain the charging strategy.

5. The battery charging protection method based on intelligent control according to claim 4 is characterized in that: The step of inputting the charging strategy into a power electronic control circuit to perform charging control and obtain a charging control result includes: Inputting the charging current-time curve, charging voltage-time curve and PWM control signal in the charging strategy into a power electronic control circuit with a digital signal processor as the core to obtain control instruction data; Performing high-speed signal processing and PWM generation on the control instruction data through a field programmable gate array to obtain a power switch control signal; Based on the power switch control signal, a full-bridge LLC resonant conversion topology structure using silicon carbide MOSFET is driven to obtain accurate charging current and charging voltage output; Performing closed-loop control on the charging current through a PI algorithm to obtain a current control loop; Performing closed-loop control on the charging voltage through a PI algorithm to obtain a voltage control loop; The current control loop and the voltage control loop are automatically switched according to the charging stage, the current control loop is used in the pre-charging and constant current stages, the voltage control loop is used in the constant voltage stage, and PWM direct control is used in the trickle current stage to obtain the charging control result.

6. The battery charging protection method based on intelligent control according to claim 1 is characterized in that: The executing a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result includes: Real-time monitoring of the charging current, charging voltage, temperature and charging status value in the battery parameter data, performing status evaluation once at a prescribed time interval to obtain abnormal status detection results; Comparing and judging the charging current with the maximum allowable charging current threshold, triggering a current abnormality alarm when the charging current is greater than the maximum allowable charging current threshold, and obtaining a current abnormality detection result; Comparing and judging the charging voltage with the charging termination voltage, triggering a voltage abnormality alarm when the charging voltage is greater than a specified proportion of the charging termination voltage, and obtaining a voltage abnormality detection result; Performing a multi-level threshold judgment on the temperature, reducing the charging current when the temperature exceeds a first threshold, further reducing the charging current when the temperature exceeds a second threshold, and immediately stopping charging when the temperature exceeds a third threshold, to obtain a temperature abnormality protection result; The current anomaly detection result, the voltage anomaly detection result and the temperature anomaly protection result are combined and processed to construct a multi-level protection mechanism of software protection layer and hardware protection layer to obtain a fault handling solution; The charging control result is adjusted based on the fault handling solution, and when a recoverable fault is detected, the charging parameters are adjusted to continue charging, and when an unrecoverable fault is detected, the charging is terminated and an alarm signal is issued to obtain the safety protection result.

7. The battery charging protection method based on intelligent control according to claim 3 is characterized in that: The step of performing reinforcement learning update on the back propagation neural network model according to the safety protection result and the charging cycle data to obtain optimized battery charging protection control parameters includes: Combining the safety protection results and the charging cycle data into a training sample set, corresponding to the battery response characteristics under different charging conditions, to obtain back propagation neural network model update data; Normalizing the update data of the back propagation neural network model to eliminate the dimension effect and obtain standardized training data; Inputting the standardized training data into the back propagation neural network model, calculating the loss function value according to the difference between the actual charging effect and the expected charging effect, and obtaining the model error; Performing gradient descent adjustment on the weight parameters of the back propagation neural network model based on the model error to obtain updated network parameters; Performing time series analysis on the battery capacity and internal resistance in the battery parameter data, calculating the capacity attenuation trend and the internal resistance change trend, and obtaining battery aging characteristic data; The updated network parameters are compensated and adjusted based on the battery aging characteristic data and applied to the prediction of the control parameters of the next charging cycle, including the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, to obtain the optimized battery charging protection control parameters.

8. A battery charging protection system based on intelligent control, characterized in that: Used to implement the battery charging protection method based on intelligent control as described in any one of claims 1 to 7, the battery charging protection system based on intelligent control comprises: The acquisition module is used to acquire the parameters of the voltage, current, temperature and internal resistance of the battery to obtain the battery parameter data, and to calculate and process the battery parameter data to obtain the charging state value, health state value and power state value; A construction module, used to construct a back propagation neural network model based on the battery parameter data, the charging state value, the health state value and the power state value to obtain an optimal charging current value, a maximum allowable charging current threshold, a charging termination voltage and a charging risk level; A generating module, configured to generate a segmented charging curve according to the optimal charging current value, the maximum allowable charging current threshold, the charging termination voltage and the charging risk level, and obtain a charging strategy; A control module, used for inputting the charging strategy into a power electronic control circuit to perform charging control and obtain a charging control result; An execution module, configured to execute a multi-level safety protection mechanism based on the battery parameter data, the charging strategy and the charging control result to obtain a safety protection result; An updating module is used to perform reinforcement learning update on the back propagation neural network model according to the safety protection results and charging cycle data to obtain optimized battery charging protection control parameters.

9. A battery charging protection device based on intelligent control, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the battery charging protection method based on intelligent control as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor is caused to execute the battery charging protection method based on intelligent control according to any one of claims 1 to 7.

Images