Intelligent Buck-Boost lithium battery charging method based on ML-MPPT and dynamic reference voltage regulation and control

Abstract

The invention relates to the technical field of lithium battery charging under photovoltaic panel power supply, in particular to a Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation, and the method comprises the steps: dividing a charging stage into a trickle charging stage, a constant-current charging stage and a constant-voltage charging stage according to the voltage of a lithium battery; the method comprises the following steps: acquiring environment temperature, solar illuminance and maximum power point voltage data in a constant current stage, preprocessing the data, training a maximum power point voltage prediction machine learning model, distilling the model to obtain a compression model, and deploying the compression model on a microcontroller; real-time environment data is input into the model to predict and obtain the maximum power point voltage as the reference voltage, the output voltage of the photovoltaic panel is sampled, and the input voltage of the Buck-Boost circuit is controlled by using the innovative PWM control system, that is, the duty ratio of PWM pulses is controlled through the microcontroller, the input impedance of the Buck-Boost circuit is adjusted, and the output voltage of the Buck-Boost circuit is adjusted. The output voltage of the photovoltaic panel, namely the input voltage of the switching power supply, is adjusted to the maximum power point voltage according to environmental conditions; and dynamically switching the lithium battery in different charging stages according to the battery state and the input power. The method integrates machine learning into prediction of the maximum power point voltage, and is suitable for an intelligent charging scene of the lithium battery under photovoltaic power supply.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery charging under photovoltaic panel power supply, and specifically to a Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation. Background Art

[0002] MPPT is a key technology in photovoltaic power generation systems. Its function is to ensure that solar cell modules can output electrical energy at optimal efficiency.

[0003] Common algorithms for MPPT technology include the perturbation and observation method (P&O algorithm), which periodically adjusts the output voltage, observes power changes, and gradually approaches the maximum power point; the incremental conductance method (IC method), which determines whether the photovoltaic cell has reached the maximum power point by calculating the incremental conductance; the constant voltage tracking method, which controls the output of the photovoltaic cell by preset voltage and is suitable for environments with uniform lighting; model predictive control, which establishes a mathematical model of the photovoltaic system and predicts changes in light intensity over a period of time in the future; and intelligent algorithms, such as artificial neural networks (ANN), genetic algorithms (GA), and fuzzy logic control (FLC), to optimize the power output of the photovoltaic system.

[0004] For example, patent CN119356475A proposes a strategy for real-time adjustment of the perturbation step size based on electrical information. This strategy dynamically adjusts the duty cycle perturbation amplitude through the power variation sensitivity coefficient, addressing the power oscillation problem of traditional perturbation-observation methods under fluctuating illumination and enabling fast and accurate maximum power point tracking of fluctuating photovoltaic arrays. The paper "Application of BP Neural Networks in Photovoltaic MPPT Control" constructs a nonlinear mapping model between temperature-irradiance input and maximum power point voltage output, enabling high-precision voltage tracking in complex environments. Patent CN119270990A employs a time-delayed averaging method to address system lags and, combined with ordinary differential equation stability analysis, establishes a quantitative relationship between controller parameters. This addresses the low tracking accuracy and speed of traditional maximum power point tracking control methods, as well as their inability to cope with complex operating environments, variations in the array's own parameters, and system lags. Patent CN119496381A proposes a dynamic reference power adjustment mechanism based on a DC-DC circuit. This mechanism generates a perturbation voltage through real-time comparison of output power with a reference power, enabling simple and rapid tracking of the photovoltaic maximum power point and implementing hardware DC-DC MPPT tracking.

[0005] In general, the goal of photovoltaic MPPT technology is to ensure that solar cell modules always operate at the maximum power operating point under different lighting conditions, thereby maximizing the power generation efficiency of the photovoltaic system.

[0006] The perturbation and observation method (P&O algorithm) is simple and easy to implement, but it has problems of oscillation and insufficient performance near the maximum power point. In the case of multiple peaks, it may be misled to the local maximum power point and cannot find the global maximum power point; the incremental conductance method (IC method) is sensitive to parameters and may fall into the local maximum power point in the case of multiple peaks; the constant voltage tracking method ignores the influence of temperature changes; the fuzzy control method is difficult to design. Although it is simple to operate, designing a good fuzzy control system requires rich experience and skills, and has high requirements for designers. The performance is affected by parameters: some parameters in fuzzy control, such as membership functions and fuzzy rules, have a great impact on system performance. Improper parameter selection may lead to system performance degradation; traditional machine learning algorithms cause MPPT delays due to calculations.

[0007] The output power and voltage curves of photovoltaic modules under different illumination are as follows: Figure 5 As shown by Figure 6 It can be seen that at the same temperature, the stronger the illumination, the greater the output power of the photovoltaic module, and the maximum power point increases with the increase of illumination; the charging process of the lithium-ion battery can be divided into four stages: trickle charging (low-voltage pre-charging), constant current charging, constant voltage charging and charging termination, such as Figure 7 As shown:

[0008] Stage 1: Trickle Charge - Trickle charge is used to pre-charge (recovery charge) a fully discharged battery cell. Trickle charge is used when the battery voltage is below 3V. The trickle charge current is one-tenth of the constant current charge current, or 0.1C (for example, if the constant charge current is 1A, the trickle charge current is 100mA).

[0009] After excessive discharge, battery performance may be affected. The main reasons for restorative charging of lithium batteries are to preserve battery life, improve battery performance, and ensure proper device operation. Improper operation during the power restoration process, such as excessive charging current or unstable voltage, can damage the battery's internal structure, further reducing performance and even shortening its service life.

[0010] Phase 2: Constant Current Charging - When the battery voltage rises above the trickle charge threshold, the charging current is increased to constant current charging. The constant current charging current ranges from 0.2C to 1.0C. The battery voltage gradually increases during the constant current charging process. This voltage is generally set between 3.0-4.2V for a single battery.

[0011] Phase 3: Constant Voltage Charging - When the battery voltage reaches 4.2V, constant current charging ends and constant voltage charging begins. The current decreases from its maximum value as the charging process continues, based on the cell's saturation level. When it reaches 0.01C, charging is considered complete. (C represents the nominal battery capacity versus the charging current. For example, for a 1000mAh battery, 1C is a charging current of 1000mA.)

[0012] Stage 4: Charge Termination - There are two typical charge termination methods: using a minimum charge current or using a timer (or a combination of the two). The minimum current method monitors the charge current during the constant voltage charge phase and terminates the charge when the charge current decreases to a range of 0.02C to 0.07C. The second method starts at the beginning of the constant voltage charge phase and terminates the charge process after two hours of continuous charging.

[0013] Current photovoltaic MPPT standards are based on current environmental conditions to ensure that solar panels can output power at optimal efficiency. They do not consider the relationship between the voltage under the photovoltaic panel MPPT and the DC-DC circuit and the charging curve of the lithium battery. The charging curve of the lithium battery (TC trickle current - CC constant current - CV constant voltage) will limit the maximum power point of the photovoltaic panel. The maximum power point (MPP) voltage V mpp and lithium battery charging voltage V bat There is a natural mismatch. Summary of the Invention

[0014] The purpose of the present invention is to provide a Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation to solve the problem that the existing photovoltaic panel MPPT technology proposed in the above background technology does not take into account the relationship between the photovoltaic panel MPPT voltage and the DC-DC circuit and the charging curve of the lithium battery.

[0015] To achieve the above object, the present invention provides the following technical solutions:

[0016] A Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control includes the following steps:

[0017] S10: Detect the lithium battery voltage and enter the corresponding charging stage according to the voltage range. When the voltage of a single lithium battery is lower than 3V, it enters the trickle charging stage; when the voltage of a single lithium battery is 3.0-4.2V, it enters the constant current charging stage; when the voltage of a single lithium battery is greater than 4.2V, it enters the constant voltage charging stage;

[0018] S20: During the constant current charging phase, the MPPT charging mode of the photovoltaic panel is started, and the temperature, illumination, and corresponding maximum power point data of all operating conditions are collected. After data preprocessing, a maximum power point voltage prediction machine learning model is trained. The model is distilled to obtain a compressed model and deployed on the microcontroller.

[0019] S30: Input the real-time collected temperature data and illumination data into the compression model to obtain the maximum power point voltage and use it as the reference voltage. The duty cycle of the PWM pulse is controlled by the microcontroller, and the input impedance of the Buck-Boost circuit is adjusted to make the photovoltaic panel voltage reach the maximum power point voltage.

[0020] S40: In each charging stage, according to the battery status and input power, dynamic switching between the MPPT mode stage and other different charging stages is achieved to ensure charging safety and efficiency.

[0021] The data preprocessing in step S20 includes abnormal data processing, missing value filling and normalization operation. The normalization operation maps the data to the range of [0, 1]. The normalization formula is:

[0022]

[0023] Where x is the data to be normalized, x(min) and x(max) represent the minimum and maximum values ​​of the data to be normalized, respectively.

[0024] Preferably, the machine learning model in step S20 includes a BP neural network, an RBF neural network, an SVM, a random forest, a decision tree or a GRU, and during model training, the preprocessed data set is divided into a ratio of 70% training set, 15% validation set and 15% test set;

[0025] Preferably, the BP neural network adopts a three-layer architecture, including an input layer, a hidden layer and an output layer. The input layer receives the ambient temperature, the light radiation illumination and the output voltage value corresponding to the maximum power point, and the output is the output voltage value corresponding to the maximum power point.

[0026] The model is evaluated by indicators such as mean absolute percentage error (MAPE), mean absolute error (MAE), mean square error (MSE), and R2 score. The model evaluation formula is as follows:

[0027]

[0028]

[0029] Where n represents the number of samples, represents the true value of the i-th sample, is the predicted value of the sample, and is the average value of all samples of the output voltage value corresponding to the maximum power point.

[0030] As a preference, the process of model distillation in step S20 is as follows: the trained converged BP neural network is used as the teacher model, a lightweight RBF network is constructed as the student model, and the ambient temperature T, light radiation illuminance RI, and the output voltage value V corresponding to the maximum power point are extracted from the teacher model. mpp , build a distillation training data set, that is, run the converged BP neural network in the training environment, collect the ambient temperature T, light radiation illuminance RI, and record the output voltage value V corresponding to the maximum power point of the BP neural network mpp , divide the data into training set and validation set; the distillation training process is to make the RBF network output approach the BP neural network through supervised learning. The formula of the MSE loss function is:

[0031]

[0032] Mini-batch SGD is used to train the RBF student model using the Adam optimizer with the learning rate set to a low value.

[0033] Preferably, in step S30, the output voltage of the photovoltaic panel is collected in real time by a voltage sampling and A / D conversion circuit, and is input into the microcontroller as a forward feedback voltage, thereby adjusting the duty cycle of the PWM pulse.

[0034] As a preference, a microcontroller with a neural network processing unit (NPU) function and a high operating frequency, such as STM32N6, is selected to complete model reasoning, data acquisition, and A / D work in real time, and the reference voltage V predicted by the machine learning model is used as the reference voltage. ref-mpp Stored in the CCRx register as the duty cycle target value, the input voltage V in That is, the output voltage of the photovoltaic panel is used as the feedback voltage, which is read by the ADC and used to dynamically adjust the duty cycle. The closed-loop control is achieved by updating the CCRx register through software.

[0035] Preferably, in step S30, the maximum power point voltage output by the compression model is denormalized, and the denormalization formula is: x=x'×(x(max)-x(min))+x(min), where x is the original data, x' is the normalized data, and x(min) and x(max) are the minimum and maximum values ​​of the data, respectively.

[0036] Preferably, the connection between MPPT and trickle charging in step S40 is as follows: when it is detected that the battery voltage is lower than 3V, the MPPT algorithm is suspended, and the system enters the trickle charging mode, charging at a current of 0.05C, and when the battery voltage returns to above 3V, it switches to the MPPT charging mode.

[0037] Preferably, the dynamic balance method of MPPT and constant current charging in step S40 includes: when there is sufficient light, the MPPT algorithm is enabled to track the maximum power point, and the output current is kept constant by adjusting the PWM duty cycle of Buck-Boost; when there is insufficient light, the boost characteristic of Buck-Boost is utilized to charge the lithium battery with a weak charging current, or an adapter or battery pack is used as a supplementary power source to maintain constant current charging.

[0038] Preferably, during the constant voltage charging stage in step S40, the MPPT algorithm stops running, and the Buck-Boost circuit reduces the charging current by adjusting the duty cycle. When the charging current decreases to a range of 0.02C to 0.07C, charging is terminated.

[0039] Preferably, the Buck-Boost circuit includes a switch tube, an inductor, a diode, a capacitor, and a PWM driver. When the switch tube is closed, the inductor is directly connected to both ends of the photovoltaic panel, the current gradually increases, and the output end relies on self-discharge to power the lithium battery; when the switch tube is turned off, the inductor supplies power to the output capacitor and the lithium battery through the diode, and the inductor volt-second conservation law is satisfied when the system is working stably.

[0040] When the power tube Q1 is closed, at the input end, the inductor L1 is directly connected to both ends of the photovoltaic panel. At this time, the inductor current gradually increases, and the di / dt increases during the turn-on transient. At the output end, the capacitor C1 relies on its own discharge to provide energy for the lithium battery.

[0041] When the power tube Q1 is turned off, since the current of the inductor L1 cannot change suddenly, the inductor supplies power to the output capacitor C1 and the lithium battery through the freewheeling tube D1;

[0042] After the system is stable, the inductor volt-second conservation law is established. When the power tube Q1 is turned on, the inductor voltage is equal to the input voltage V in When the power tube Q1 is turned off, the inductor voltage is equal to the output voltage V out ; Let T be the period, T on is the on-time, T off is the off time, D is the duty cycle (D=T on / T);

[0043] According to the conservation of inductance volt-second:

[0044] V in ·T on =V out ·T off ,

[0045] Vin·D·T=Vout·(1-D)·T,

[0046] From this we can get:

[0047] Vout=D / (1-D)·Vin,

[0048] D=Vout / (Vout+Vin),

[0049] When the duty cycle is less than 0.5, the output voltage is reduced; when the duty cycle is greater than 0.5, the output voltage is increased.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation provided by the present invention combines machine learning maximum power point tracking (ML-MPPT) technology with dynamic reference voltage regulation and relies on a Buck-Boost circuit to achieve intelligent charging of lithium batteries. It has the following beneficial effects:

[0052] 1. Improving Charging Efficiency: The ML-MPPT mode is adopted during the constant current charging stage. By training the machine learning model and obtaining a compressed model through distillation, the maximum power point voltage can be quickly and accurately predicted based on real-time temperature and illumination data. This is used as the reference voltage to regulate the input impedance of the Buck-Boost circuit, ensuring that the photovoltaic panel always operates at the maximum power point, significantly improving the utilization rate of photovoltaic energy, and thus improving the overall charging efficiency.

[0053] 2. Realize intelligent dynamic switching: According to the range of lithium battery voltage, it is divided into trickle current, constant current and constant voltage charging stages. According to the battery status and input power, it realizes dynamic switching between the MPPT constant current charging stage and other charging stages. It can not only efficiently utilize photovoltaic energy when there is sufficient sunlight, but also ensure that the charging process matches the charging curve characteristics of the battery in different charging stages, taking into account both charging efficiency and safety.

[0054] 3. Optimize model deployment and operation: Distill the trained machine learning model to obtain a compressed model, significantly reducing model parameters and lowering the usage of hardware resources such as microcontrollers. This ensures that the model can be stably and efficiently deployed on embedded devices while maintaining high prediction accuracy to meet real-time charging control requirements.

[0055] 4. Ensure charging safety: Targeted control measures are taken at each charging stage. For example, charging with a low current in the trickle stage to avoid battery damage, disabling the MPPT in the constant voltage stage and dynamically adjusting circuit parameters to stabilize the voltage within a safe range, and adjusting the PWM duty cycle in real time through voltage sampling feedback to effectively prevent overcharging, overvoltage and other problems, thereby extending battery life.

[0056] 5. Enhanced adaptability and flexibility: The method uses a variety of machine learning models (such as BP neural network, RBF neural network, etc.) to adapt to data analysis needs under different working conditions; the Buck-Boost circuit can achieve buck-boost function through duty cycle adjustment, adapting to the matching requirements of photovoltaic panel output voltage and lithium battery charging voltage, thereby improving the applicability of the method in complex environments and different equipment configurations. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they provide further detailed explanation, but do not constitute a limitation of the present invention.

[0058] Figure 1 This is a diagram of the lithium battery charging technology architecture that integrates machine learning MPPT and PWM in the present invention;

[0059] Figure 2 This is the BP neural network architecture diagram of the present invention;

[0060] Figure 3 The BP neural network of the present invention is distilled into the RBF network architecture diagram;

[0061] Figure 4 This is a flow chart of the present invention using the BUCK-BOOST buck-boost topology for charging;

[0062] Figure 5 Draw IV curves for photovoltaic modules at different temperatures;

[0063] Figure 6 Create PV curves for photovoltaic panels at different temperatures;

[0064] Figure 7 This is the charging curve of the lithium battery. DETAILED DESCRIPTION

[0065] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation of the present invention comprises a Buck-Boost circuit, an MPPT control module, and trickle charging, constant current charging, and constant voltage charging modules for the lithium battery.

[0067] The lithium battery charging technology solution proposed by the present invention that integrates machine learning MPPT and PWM is as follows Figure 1 As shown in the figure, the BUCK-BOOST circuit included in this technical solution is composed of a switch tube Q1, an inductor L1, a diode D1, a capacitor C1 and a PWM generator; the MPPT control circuit is composed of a machine learning model for predicting the MPPT voltage, a photovoltaic panel output voltage (V) sampling and A / D conversion circuit, a temperature (℃) sampling and A / D conversion circuit, and a solar irradiance (W / m 2 ) sampling and A / D conversion circuit and MCU; the efficient charging and safety protection of lithium batteries are composed of trickle current, constant current and constant voltage control modules.

[0068] When the PWM controller detects that the voltage of a single-cell lithium battery is lower than about 3V, trickle charging is adopted; when the voltage of a single-cell lithium battery is 3.0-4.2V, constant current charging is adopted; when the voltage of a single-cell lithium battery is greater than 4.2V, constant voltage charging is adopted.

[0069] During the constant current charging phase of the lithium battery, the photovoltaic panel MPPT charging mode is started. The temperature, illumination and corresponding maximum power point data of the full working conditions are collected, and abnormal data processing and missing value filling are performed. The machine learning model is trained, and finally a machine learning model such as SVM, BP neural network, and RBF is obtained. The machine learning model is distilled to obtain a distilled compressed model to adapt to running the model on a microcontroller (MCU) or other resource-constrained hardware platform. The temperature data and illumination data are input into the distillation model to obtain the maximum power point voltage V mpp and the maximum power point voltage V mpp As the reference voltage. The voltage of the photovoltaic panel is used as the forward feedback voltage and input into the control circuit with MCU as the core to control the duty cycle of the PWM pulse and adjust the input impedance of the BUCK-BOOST circuit, so that the photovoltaic panel voltage quickly reaches the maximum power point voltage V mpp .

[0070] 1. BUCK-BOOST module

[0071] The main factors affecting photovoltaic power generation include light radiation illumination and ambient temperature. Solar irradiance is not continuous. For example, it may be blocked by buildings and clouds, resulting in insufficient sunlight. There are also differences in irradiance between morning, noon and evening. To adapt to complex working conditions, the buck-boost circuit's buck-boost adaptability is used to provide sufficient charging power for lithium batteries.

[0072] When power transistor Q1 is closed, inductor L1 is directly connected to the photovoltaic panel at the input, causing the inductor current to gradually increase. During the turn-on transient, di / dt increases, and at the output, C1 provides energy to the lithium battery through its own discharge.

[0073] When the power tube Q1 is turned off, since the current of the inductor L1 cannot change suddenly, the inductor supplies power to the output capacitor C1 and the lithium battery through the freewheeling tube D1.

[0074] After the system is stable, the inductor volt-second conservation is achieved. When Q1 is turned on, the inductor voltage is equal to the input voltage V in When Q1 is turned off, the inductor voltage is equal to the output voltage V out . Let T be the period, T on is the on-time, T off is the off time, D is the duty cycle (D=T on / T).

[0075] According to the conservation of inductance volt-second:

[0076] V in ·T on =V out ·T off ,

[0077] Vin·D·T=Vout·(1-D)·T,

[0078] From this we can get:

[0079] Vout=D / (1-D)·Vin,

[0080] D=Vout / (Vout+Vin),

[0081] When the duty cycle is less than 0.5, the output voltage is reduced; when the duty cycle is greater than 0.5, the output voltage is increased.

[0082] 2. MPPT control module

[0083] The MPPT control module consists of a voltage sampling and A / D module, an MPPT machine learning model and an MCU module. The voltage sampling and A / D module samples the output voltage of the photovoltaic panel and performs digital-to-analog conversion. The solar irradiance and ambient temperature are input into the MPPT machine learning model to predict the MPPT voltage of the photovoltaic panel. The MPPT voltage is used as the reference voltage of the PWM generator. The duty cycle is adjusted in real time through PWM. The output voltage of the photovoltaic panel quickly tracks the MPPT voltage, realizing the tracking of the maximum power point of the photovoltaic panel.

[0084] 2.1 Voltage sampling and A / D module

[0085] The voltage sampling and A / D conversion modules of the MCU are used to collect the output voltage of the photovoltaic panel in real time and perform digital-to-analog conversion.

[0086] 2.2 MPPT Machine Learning Model

[0087] 2.2.1 Data Collection

[0088] Data collection includes ambient temperature, solar irradiance and output voltage of photovoltaic panels. Irradiance can be sampled using mature integrated modules such as BH1750FVI.

[0089] 2.2.2 Data Preprocessing

[0090] Fill missing values ​​in the collected photovoltaic panel data, process abnormal data, and normalize the data set to scale features and labels of different scales to the range of [0, 1] to eliminate the impact of dimension:

[0091]

[0092] Among them, x is the data to be normalized, x(max) and x(min) represent the maximum and minimum values ​​in the data to be normalized, respectively. Finally, after the data is mapped between 0 and 1, it is input into the model as feature data for training.

[0093] The preprocessed dataset is divided into training, validation, and test sets for use during training. A common ratio is 70% training, 15% validation, and 15% test.

[0094] 2.2.3 Model Construction and Evaluation

[0095] The MPPT machine learning model can be built using machine learning algorithms such as BP, RBF, SVM, RF (random forest), DT (decision tree) and GRU. BP, RBF, SVM, RF (random forest), DT (decision tree) and GRU machine learning algorithms are preferred. Here, the BP neural network is used as an example. The BP neural network adopts a three-layer architecture, including an input layer, a hidden layer and an output layer. The input layer receives the ambient temperature, light radiation illuminance and the output voltage value corresponding to the maximum power point, and the output is the output voltage value corresponding to the maximum power point. The model is evaluated by indicators such as the mean absolute percentage error (MAPE), the mean absolute error (MAE), the mean square error (MSE), and the R2 score. The model evaluation formula is as follows:

[0096]

[0097] Where n represents the number of samples, y i represents the true value of the i-th sample, is the predicted value of the sample; is the average of all samples of the output voltage corresponding to the maximum power point. Trained BP, RBF, SVM, RF (random forest), DT (decision tree), and GRU models require significant computing resources and storage space. These models need to be compressed and optimized to fit embedded devices with limited memory, computing power, and energy consumption. This reduces model size and computational requirements without compromising performance.

[0098] 2.2.4 Model Distillation

[0099] Model distillation is a method of model compression. After the trained BP network converges, the BP network is distilled into an RBF network.

[0100] After the BP neural network is trained and converged, a lightweight RBF network is constructed as a teacher model and a student model is constructed. The ambient temperature T, light radiation illuminance RI, and the output voltage value V corresponding to the maximum power point are extracted from the teacher model. mpp , build a distillation training data set, that is, run the converged BP neural network in the training environment, collect the ambient temperature T, light radiation illuminance RI, and record the output voltage value V corresponding to the maximum power point of the BP neural network mpp , divide the data into training set and validation set; the distillation training process is to make the RBF network output close to the BP neural network action through supervised learning, with MSE as the loss function, Mini-batch SGD is used to train the RBF student model using the Adam optimizer with the learning rate set to a low value.

[0101] Through this distillation process, the RBF network inherits the control strategy of the BP neural network while significantly reducing computational complexity, meeting the real-time control requirements of the power converter. This method, while retaining the performance advantages of the BP neural network, addresses the resource bottleneck issue of the BP neural network in embedded deployment.

[0102] 2.2.5 Model Deployment

[0103] To minimize model size and computational complexity, TinyML technology is used to run machine learning models on microcontrollers (MCUs) or other resource-constrained hardware platforms, extending AI technology to Internet of Things (IoT) devices, giving them intelligent perception and processing capabilities, thereby enabling wider applications.

[0104] 2.2.6 Model Application

[0105] The real-time measured ambient temperature T and light radiation illuminance RI are input into the RBF model to perform V mpp Real-time prediction and anti-normalization operation are performed.

[0106] x=x'×(x(max)-x(min))+x(min),

[0107] Where x is the original data, x′ is the normalized data, and x(min) and x(max) are the minimum and maximum values ​​of the data, respectively.

[0108] Denormalization converts the normalized data back to its original scale. The specific steps depend on the normalization method used during preprocessing. Be sure to save the normalization parameters during training and use them during denormalization to obtain predictions at the original scale.

[0109] 2.2.7. Generate PWM signal

[0110] The real-time monitored ambient temperature T and light radiation illuminance RI are fed into the rbf model to predict the MPPT voltage V mpp , V mpp As a reference voltage, the output voltage of the photovoltaic panel tracks V mpp , and get the maximum power point voltage.

[0111] 3. Lithium battery trickle current, constant current and constant voltage charging modules

[0112] This device uses a buck-boost topology for charging, automatically adjusting the operating mode when the battery voltage is below, above, or equal to the input voltage, ensuring the battery is always optimally charged. During the charging process, the chip supports multiple stages: trickle charge (TC), constant current charge (CC), constant voltage charge (CV), and charge termination.

[0113] 3.1. Connection between MPPT and trickle charging

[0114] When the MCU detects that the battery voltage is below a safety threshold (e.g., a single-cell lithium battery voltage <3.0V), the MPPT algorithm suspends to prevent high-voltage input from impacting the battery. The charging system is forced into trickle mode, using a low current (typically 0.05C) to restore battery activity and prevent damage from deep discharge. Once the battery voltage rises above the trickle threshold, the system switches to MPPT charging mode, using the buck-boost circuit to adjust the input voltage and current to maximize the solar panel's output power.

[0115] 3.2 Dynamic balance between MPPT and constant current charging

[0116] In the constant current stage, the MPPT target power upper limit is P max =I charge ·V bat, it is necessary to dynamically adjust the PWM duty cycle to make the photovoltaic output power approach this value.

[0117] 3.2.1. When there is sufficient light

[0118] When the battery voltage returns to normal and the input source is a photovoltaic module, the MPPT algorithm is enabled to track the maximum power point. By adjusting the BUCK-BOOST PWM duty cycle to optimize energy conversion efficiency, the BUCK-BOOST circuit converts the input power into a constant current output (such as 1C), rapidly increasing the battery voltage.

[0119] When the charging current in MPPT mode exceeds the battery's allowed threshold (such as 1C), it automatically switches to constant current charging to protect battery life.

[0120] 3.2.2 When there is insufficient light

[0121] If the input power cannot maintain the constant current requirement, the Buck-Boost circuit utilizes the boost characteristic to charge the lithium battery at a weak charging current, or uses an adapter or battery pack as a supplemental power source to maintain constant current charging. In dual-input scenarios (such as solar power and mains power), the Buck-Boost circuit prioritizes solar MPPT power supply, with mains power as a backup supplement, to maximize energy efficiency.

[0122] 3.3 Power Adaptation in Constant Voltage Charging Phase

[0123] When the battery voltage approaches saturation (such as 4.2V for a lithium battery), the system switches to constant voltage mode and the MPPT algorithm stops operating. The buck-boost circuit reduces the charge current by adjusting the duty cycle to prevent overvoltage risks.

[0124] 3.4 Charging termination

[0125] Monitor the charging current during the constant voltage charging phase and terminate the charging when the charging current decreases to the range of 0.02C to 0.07C.

[0126] When the battery temperature or voltage is abnormal, the MPPT is immediately interrupted and charging is terminated.

[0127] Through this mechanism, the BUCK-BOOST charging circuit achieves seamless switching among trickle current, constant current, and constant voltage stages, fully utilizing renewable energy while ensuring the safety and efficiency of the battery charging process.

[0128] Example 1: Smart Charging Solution Based on BP Neural Network and RBF Distillation Model

[0129] Experimental conditions:

[0130] Photovoltaic panel parameters: maximum power 200W, open circuit voltage 36V, short circuit current 6.8A;

[0131] Lithium battery: single 3.7V / 10Ah lithium iron phosphate battery, connected in series to form a 12V battery pack;

[0132] Environmental conditions: temperature range -10~50℃, irradiance 200~1000W / m 2 (Simulates cloudy to sunny weather).

[0133] Technical details:

[0134] Data collection: 5000 sets of temperature, irradiance and corresponding MPP voltage data were collected. After outliers were removed (3σ criterion) and missing values ​​were filled by linear interpolation, the data were divided into training set, validation set and test set with a ratio of 70%:15%:15%.

[0135] Model training: A three-layer BP neural network (2 neurons in the input layer, 10 neurons in the hidden layer, and 1 neuron in the output layer) was used, with MSE as the loss function and Adam optimizer (learning rate 0.001) trained until convergence (MAPE < 2%).

[0136] Model distillation: The BP network is used as the teacher model, and a two-layer RBF network (with 2 neurons in the input layer and 5 neurons in the hidden layer) is constructed as the student model. Distillation training is performed for 500 rounds (learning rate 0.0005). After compression, the model parameters are reduced by 60%.

[0137] Charging control: ML-MPPT is enabled during the constant current stage (3.0-4.2V), and real-time temperature (DS18B20 sensor) and irradiance (BH1750FVI module) are collected. The reference voltage is output through the RBF model, and the PWM duty cycle is adjusted in steps of 0.1%.

[0138] Experimental results:

[0139] When the light intensity fluctuates (e.g. from 500 to 800W / m 2 Mutation), MPP tracking response time <0.5s, power fluctuation amplitude <3%;

[0140] The overall charging efficiency (photovoltaic panel output power / battery absorbed power) reaches 92.3%, an increase of 8.5% compared to the traditional P&O algorithm.

[0141] Example 2: Dual-input source (solar + mains) collaborative charging solution

[0142] Experimental conditions:

[0143] New hardware: 220V to 12V adapter (power 100W) as backup power supply, input switched via relay;

[0144] Lithium battery: single 4.2V / 20Ah ternary lithium battery, connected in series to form a 24V battery pack;

[0145] Environmental conditions: simulate weak light in the morning and evening (irradiance < 300W / m 2 ) and strong sunlight at noon (irradiance> 800W / m 2 ).

[0146] Technical details:

[0147] Mode switching logic:

[0148] When the photovoltaic MPPT output power is greater than or equal to the constant current requirement (1C=20A), only solar energy is used for charging;

[0149] When the power is less than 0.7C (14A), it switches to mains power supply (PV + mains power combined power supply);

[0150] Safety control: During the trickle current stage (<3.0V), the battery is charged at 0.05C (1A). During the constant voltage stage (>4.2V), the MPPT is turned off and the PWM duty cycle is dynamically adjusted to a stable output voltage of 4.2V±0.02V.

[0151] Experimental results:

[0152] In low light environment (200W / m 2 ), charging time is shortened by 40% compared with pure solar energy solution;

[0153] There is no overcharging throughout the process (voltage fluctuation <0.05V), and the battery cycle life test (1000 times) has a capacity retention rate of ≥85%.

[0154] Example 3: Optimization of extreme environment adaptability

[0155] Experimental conditions:

[0156] Environmental simulation: high temperature (60°C), low temperature (-20°C) and rapid shading (irradiance from 1000 to 200W / m within 10s 2 );

[0157] Lithium battery: single-cell 3.2V / 5Ah lithium titanate battery (wide temperature characteristics).

[0158] Technical details:

[0159] Model optimization: 2,000 sets of extreme temperature samples (-20 to 60°C) were added to the training data. A GRU network (with a 5s time window) was used to predict irradiance trends, and the reference voltage was adjusted 50ms in advance.

[0160] Circuit adaptation: The Buck-Boost inductor uses a ferrite core with a temperature resistance of 125°C, and the switching tube uses SiC MOSFET (with a voltage resistance of 600V) to reduce high-temperature losses.

[0161] Experimental results:

[0162] MPPT tracking accuracy (measured MPP voltage / theoretical value) at extreme temperatures is ≥98%, while traditional IC algorithms are only 90%;

[0163] During rapid shading, the power drop is less than 5%, the recovery time is less than 0.3s, and there is no obvious oscillation.

[0164] Comparative Example 1: Traditional Perturbation and Observation (P&O) Charging Solution

[0165] Experimental conditions: Same as Example 1, but the MPPT algorithm was replaced by a fixed step size (0.5V) perturbation observation method.

[0166] Technical defects:

[0167] When the light intensity fluctuates (e.g. from 800 to 500W / m 2 ), the power oscillation amplitude reaches 15%, and the tracking time is > 2s;

[0168] The constant current stage cannot dynamically match the battery voltage requirement, and the charging efficiency is only 83.8%, which is 8.5% lower than that of Example 1;

[0169] It is easy to fall into local MPP when there are multi-peak illumination (partial shading), and the power loss can reach 12%.

[0170] Comparative Example 2: ML-MPPT Scheme without Model Distillation

[0171] Experimental conditions: Same as Example 1, but directly deploying an undistilled BP neural network (parameter size 1.2MB).

[0172] Technical defects:

[0173] The microcontroller (STM32N6, 64KB RAM) frequently overflows during operation, with response delays greater than 1s.

[0174] The model inference power consumption reaches 80mA, which is 300% higher than the distilled RBF model (20mA), and does not meet the low power consumption requirement;

[0175] The voltage regulation overshoot in the constant current stage reaches 5%, posing a risk of battery overvoltage.

[0176] Comparison table of key parameters and results of the above three embodiments and two comparative examples:

[0177]

[0178] From the comparison of key data between the above embodiment and the comparative example, the specific analysis is as follows:

[0179] 1. Impact of Core Algorithms on Performance

[0180] Comparison between ML-MPPT and traditional algorithms

[0181] Examples 1-3 all use machine learning (ML) combined with model distillation technology MPPT solutions, while Comparative Example 1 uses the traditional perturbation and observation (P&O) method. The data shows:

[0182] Tracking speed: The MPP tracking response time of Example 1 (0.3-0.5s) is much faster than that of Comparative Example 1 (>2s). This is because the ML model can directly predict the maximum power point (MPP) through historical data training, without the need for repeated perturbation trial and error like the P&O algorithm, thus reducing adjustment delays.

[0183] Power stability: The power fluctuation amplitude of the embodiment (<3%-5%) is significantly lower than that of comparative example 1 (15%), especially in scenarios with sudden changes in illumination (such as rapid occlusion in embodiment 3). The ML model adjusts the reference voltage in advance through timing prediction (such as the GRU network), avoiding the oscillation problem of the P&O algorithm.

[0184] Efficiency advantage: The charging efficiency of the embodiment (90.8%-92.3%) is 7%-8.5% higher than that of comparative example 1 (83.8%). The core reason is that the ML model can accurately lock the global MPP under complex lighting conditions (such as multi-peak shadows), while P&O is prone to falling into local optimality (power loss of up to 12%).

[0185] The necessity of model distillation

[0186] Examples 1-3 all used model distillation (e.g., BP neural network distillation to lightweight RBF network), while Comparative Example 2 directly used the undistilled BP model. The results showed that:

[0187] Resource adaptability: After distillation, the model parameters are reduced by 60% (Example 1), which can adapt to the limited memory (64KB) of microcontrollers (such as STM32N6). However, the undistilled model (1.2MB) of Comparative Example 2 exceeds the memory limit, resulting in response delays (>1s) and operation overflows, and cannot work stably.

[0188] Safety: The voltage fluctuations of Examples 1-3 are all less than 0.05V, and there is no overcharging risk. However, in Comparative Example 2, due to model delay and adjustment lag, the voltage overshoot reaches 5%, posing a serious overcharging risk.

[0189] 2. Scenario Adaptability Analysis

[0190] Conventional environment (Example 1)

[0191] At a temperature of -10 to 50°C and an irradiance of 200 to 1000 W / m 2 In the typical scenario, the comprehensive performance of Example 1 is the best:

[0192] The efficiency (92.3%) is the highest among the three groups of embodiments, which is due to the accurate fitting of the BP+RBF model to the conventional working conditions;

[0193] The power fluctuation is less than 3%, indicating that the regulation accuracy of the model under stable illumination is higher than that of the dual input source (Example 2, less than 5%) and extreme environment solution (Example 3, less than 5%).

[0194] Weak light and dual-source collaboration (Example 2)

[0195] For irradiance <300W / m 2 In low-light scenarios, Example 2 solves the problem of low single-source charging efficiency by synergizing "photovoltaic + mains power" dual inputs:

[0196] The charging time under weak light conditions is 40% shorter than that of the pure solar solution, while the efficiency remains at 91.5% (only 0.8% lower than Example 1);

[0197] After 1000 cycles, the capacity retention rate is ≥85%, proving that the dual-source switching logic (photovoltaic priority, AC power supplement) does not affect the battery life, which is better than the overcharging risk caused by the voltage fluctuation (±0.1V) in Comparative Example 1.

[0198] Extreme Environment (Example 3)

[0199] In the -20 to 60°C and rapid occlusion scenarios, the GRU network of Example 3 (including extreme sample training) showed strong adaptability:

[0200] The MPP tracking response time is less than 0.3s, which is the fastest among all solutions. This is because the GRU's timing prediction capability can adjust the reference voltage 50ms in advance.

[0201] The tracking accuracy under extreme temperatures is ≥98%, which is much higher than the 90% in Comparative Example 1, indicating that the model optimization for extreme data (sample expansion of 2000 groups) effectively offsets the impact of temperature on the characteristics of photovoltaic panels.

[0202] 3. Model Resource Consumption and Hardware Adaptability

[0203] The key role of distillation technology

[0204] The model of Example 1-2 has 60% fewer parameters after distillation and can be directly deployed on a microcontroller (such as STM32N6). However, the model of Comparative Example 2, which is not distilled (with 1.2MB of parameters), exceeds the hardware memory (64KB), resulting in overflow and delay (>1s). This shows that:

[0205] Model distillation is a core technology for balancing prediction accuracy and hardware resources. Examples 1-2 maintain high performance even after parameter compression, validating the practicality of the "teacher-student" model architecture.

[0206] In Example 3, the parameter scale is increased by 20% to adapt to extreme environments, but it is still controlled within the range that the microcontroller can handle, which reflects the flexibility of the algorithm design.

[0207] Limitations of Model-Free Approaches

[0208] Comparative Example 1 (P&O algorithm) does not require model resources but relies on hardware logic to implement disturbance regulation, resulting in:

[0209] In the case of multi-peak illumination (partial occlusion), the global MPP cannot be identified, and the power loss reaches 12%;

[0210] The voltage fluctuation is ±0.1V, which poses an overcharging risk. However, the embodiment controls the fluctuation to <0.05V through dynamic reference voltage regulation of the ML model, which is safer.

[0211] 4. Battery Protection and Life Performance

[0212] Voltage stability

[0213] The voltage fluctuations of Examples 1-3 are all less than 0.05V, and there is no overcharge record, while the voltage fluctuation of Comparative Example 1 is ±0.1V and the voltage overshoot of Comparative Example 2 is 5%, indicating that:

[0214] The ML model accurately controls the PWM duty cycle of the Buck-Boost circuit by outputting a reference voltage in real time, avoiding the regulation lag of traditional algorithms.

[0215] The dynamic switching logic (such as turning off the MPPT during the constant voltage stage in Example 2) further ensures voltage stability at the end of charging.

[0216] Cycle life

[0217] The 1000-cycle capacity retention rate of Example 2 is ≥85%, which is better than the industry average (about 80%), proving that:

[0218] The refined division of charging stages (tricky current 0.05C to constant current 1C to constant voltage 4.2V±0.02V) effectively reduces battery loss;

[0219] The smooth transition (without current shock) during dual-source switching reduces damage to the electrode structure, while the fixed step adjustment in comparative example 1 easily generates current fluctuations and accelerates attenuation.

[0220] V. Summary

[0221] Technical advantages: The ML-MPPT of the present invention, combined with dynamic reference voltage regulation, outperforms traditional P&O algorithms and undistilled model solutions in tracking speed (<0.5s), efficiency (90.8%-92.3%), and safety (fluctuation <0.05V).

[0222] Scenario adaptability: Through algorithm optimization (such as GRU timing prediction) and hardware collaboration (dual input sources), it can meet the diverse needs of conventional, low-light, and extreme environments;

[0223] Engineering Value: Model distillation technology solves the resource limitations of embedded deployment, making the method feasible for practical application and providing an efficient and reliable solution for smart charging of lithium batteries.

[0224] The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage regulation provided by the present invention combines machine learning maximum power point tracking (ML-MPPT) technology with dynamic reference voltage regulation and relies on a Buck-Boost circuit to achieve intelligent charging of lithium batteries. It has the following beneficial effects:

[0225] 1. Improving Charging Efficiency: The ML-MPPT mode is adopted during the constant current charging stage. By training the machine learning model and obtaining a compressed model through distillation, the maximum power point voltage can be quickly and accurately predicted based on real-time temperature and illumination data. This is used as the reference voltage to regulate the input impedance of the Buck-Boost circuit, ensuring that the photovoltaic panel always operates near the maximum power point, significantly improving photovoltaic energy utilization and thereby improving overall charging efficiency.

[0226] 2. Realize intelligent dynamic switching: Divide the charging stages into trickle current, constant current, and constant voltage according to the range of lithium battery voltage, and realize dynamic switching between MPPT mode and each charging stage based on the battery status and input power. It can not only efficiently utilize photovoltaic energy when there is sufficient sunlight, but also ensure that the charging process matches the battery characteristics at different charging stages, taking into account both charging efficiency and safety.

[0227] 3. Optimize model deployment and operation: Distill the trained machine learning model to obtain a compressed model, significantly reducing model parameters and lowering the usage of hardware resources such as microcontrollers. This ensures that the model can be stably and efficiently deployed on embedded devices while maintaining high prediction accuracy to meet real-time charging control requirements.

[0228] 4. Ensure charging safety: Targeted control measures are taken at each charging stage. For example, charging with a low current in the trickle stage to avoid battery damage, disabling the MPPT in the constant voltage stage and dynamically adjusting circuit parameters to stabilize the voltage within a safe range, and adjusting the PWM duty cycle in real time through voltage sampling feedback to effectively prevent overcharging, overvoltage and other problems, thereby extending battery life.

[0229] 5. Enhanced adaptability and flexibility: The method uses a variety of machine learning models (such as BP neural network, RBF neural network, etc.) to adapt to data analysis needs under different working conditions; the Buck-Boost circuit can achieve buck-boost function through duty cycle adjustment, adapting to the matching requirements of photovoltaic panel output voltage and lithium battery charging voltage, thereby improving the applicability of the method in complex environments and different equipment configurations.

[0230] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control, characterized by: The following steps are involved: S10: Detect the lithium battery voltage and enter the corresponding charging stage according to the voltage range. When the voltage of a single lithium battery is lower than 3V, it enters the trickle charging stage; when the voltage of a single lithium battery is 3.0-4.2V, it enters the constant current charging stage; when the voltage of a single lithium battery is greater than 4.2V, it enters the constant voltage charging stage; S20: During the constant current charging phase, the MPPT charging mode of the photovoltaic panel is started, and the temperature, illumination, and corresponding maximum power point data of all working conditions are collected. After data preprocessing, a machine learning model is trained, and the model is distilled to obtain a compressed model and deployed on the microcontroller. S30: Input the real-time collected temperature and illumination data into the compression model to obtain the maximum power point voltage and use it as the reference voltage. The output voltage of the photovoltaic panel is collected in real time and input into the microcontroller as the forward feedback voltage. The innovative PWM control system is used to control the input voltage of the Buck-Boost circuit. That is, the microcontroller controls the duty cycle of the PWM pulse and adjusts the input impedance of the Buck-Boost circuit so that the output voltage of the photovoltaic panel, i.e., the input voltage of the switching power supply, is adjusted to the maximum power point voltage according to the environmental conditions. S40: In each charging stage, according to the battery status and input power, dynamic switching between the MPPT mode stage and other charging stages is achieved to ensure charging safety and efficiency. The MPPT mode stage is a constant current stage. The data preprocessing in step S20 includes abnormal data processing, missing value filling and normalization operation. The normalization operation maps the data to the range of [0, 1]. The normalization formula is: Where x is the data to be normalized, x(min) and x(max) represent the minimum and maximum values ​​of the data to be normalized, respectively.

2. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: The machine learning model in step S20 includes BP neural network, RBF neural network, SVM, random forest, decision tree or GRU. During model training, the preprocessed data set is divided into a ratio of 70% training set, 15% validation set and 15% test set.

3. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 2, characterized in that: The BP neural network adopts a three-layer architecture, including an input layer, a hidden layer and an output layer. The input layer receives the ambient temperature and light radiation illumination, and the output is the output voltage value corresponding to the maximum power point; By mean absolute percentage error MAPE, mean absolute error MAE, mean square error MSE, R 2 The scoring indicators are used to evaluate the model. The above scoring indicator models are V mpp_mape 、V mpp_mae 、V mpp_mse 、V mpp_R2 , the model evaluation formula is as follows: Where n represents the number of samples, represents the true value of the i-th sample, is the predicted value of the sample, and is the average value of all samples of the output voltage value corresponding to the maximum power point.

4. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: The process of model distillation in step S20 is as follows: the trained converged BP neural network is used as the teacher model, a lightweight RBF network is constructed as the student model, and the ambient temperature T, light radiation illumination RI, and the output voltage value V corresponding to the maximum power point are extracted from the teacher model. mpp , build a distillation training data set, that is, run the converged BP neural network in the training environment, collect the ambient temperature T, light radiation illuminance RI, and record the output voltage value V corresponding to the maximum power point of the BP neural network mpp , divide the data into training set and validation set; the distillation training process is to make the RBF network output approach the BP neural network through supervised learning. The formula of the MSE loss function is: Mini-batch SGD is used to train the RBF student model using the Adam optimizer with the learning rate set to a low value.

5. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: In step S30, the output voltage of the photovoltaic panel, i.e., the input voltage of the switching power supply, is collected in real time through the voltage sampling and A / D conversion circuit, and is input into the microcontroller as a forward feedback voltage, and compared with the reference voltage predicted by the machine learning model, thereby adjusting the duty cycle of the PWM pulse.

6. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: In step S30, the maximum power point voltage output by the compression model is denormalized, and the denormalization formula is: x=x'×(x(max)-x(min))+x(min), where x is the original data, x' is the normalized data, and x(min) and x(max) are the minimum and maximum values ​​of the data, respectively.

7. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: The connection between MPPT and trickle charging in step S40 is as follows: when it is detected that the battery voltage is lower than 3V, the MPPT algorithm is suspended, and the system enters the trickle charging mode, charging at a current of 0.05C. When the battery voltage rises back to above 3V, it switches to the MPPT charging mode.

8. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: The dynamic balance between MPPT and constant current charging in step S40 includes: when there is sufficient sunlight, the MPPT algorithm is enabled to track the maximum power point, and the output current is kept constant by adjusting the PWM duty cycle of Buck-Boost; when there is insufficient sunlight, the boost characteristic of Buck-Boost is utilized to charge the lithium battery with a weak charging current, or an adapter or battery pack is used as a supplementary power source to maintain constant current charging.

9. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: During the constant voltage charging phase in step S40, the MPPT algorithm stops running, and the Buck-Boost circuit reduces the charging current by adjusting the duty cycle. When the charging current decreases to a range of 0.02C to 0.07C, charging is terminated.

10. The Buck-Boost lithium battery intelligent charging method based on ML-MPPT and dynamic reference voltage control according to claim 1, characterized in that: The Buck-Boost circuit includes a switch tube Q1, an inductor L1, a diode D1, a capacitor C1, and a PWM generator. When the switch tube is closed, the inductor is directly connected to the two ends of the photovoltaic panel, the current gradually increases, and the output end relies on its own discharge to power the lithium battery; when the switch tube is turned off, the inductor supplies power to the output capacitor and the lithium battery through the diode, and when the system is working stably, the inductor volt-second conservation law is satisfied. When the power tube Q1 is closed, at the input end, the inductor L1 is directly connected to both ends of the photovoltaic panel. At this time, the inductor current gradually increases, and the di / dt increases during the turn-on transient. At the output end, the capacitor C1 relies on its own discharge to provide energy for the lithium battery. When the power tube Q1 is turned off, since the current of the inductor L1 cannot change suddenly, the inductor supplies power to the output capacitor C1 and the lithium battery through the freewheeling tube D1; After the system is stable, the inductor volt-second conservation law is established. When the power tube Q1 is turned on, the inductor voltage is equal to the input voltage V in When the power tube Q1 is turned off, the inductor voltage is equal to the output voltage V out ; Let T be the period, T on is the on-time, T off is the off time, D is the duty cycle (D=T on / T); According to the conservation of inductance volt-second: V in ·T on =V out ·T off , Vin·D·T=Vout·(1-D)·T, From this we can get: Vout=D / (1-D)·Vin, D=Vout / (Vout+Vin), When the duty cycle is less than 0.5, the output voltage is reduced; when the duty cycle is greater than 0.5, the output voltage is increased.

Images