A fast charging method for camera intelligent battery

Abstract

The present invention discloses a fast charging method for a camera intelligent battery, which relates to the technical field of camera battery charging, including generating a battery loss coefficient; inputting the battery loss coefficient into a pre-trained neural network model, and outputting a charging power upper limit value adapted to the current battery state; matching the temperature distribution data with the charging power upper limit value to generate a first round of dynamic charging power curve; in the execution stage of the first round of dynamic charging power curve, correcting the first round of dynamic charging power curve according to the trend characteristics of the voltage response curve; dividing the corrected charging power curve into multiple time periods according to a preset time interval, and charging the battery according to the power value corresponding to each time period until charging is completed. The present invention introduces a neural network model to establish a mapping relationship between the battery loss coefficient and the safe charging power, thereby realizing intelligent adjustment of charging parameters.

Description

Technical Field

[0001] The invention relates to the technical field of camera battery charging, in particular to a fast charging method for a camera intelligent battery. Background Art

[0002] With the rapid development of professional video production and live broadcasting industries, the frequency and duration of use of cameras as core equipment are constantly increasing, and the demand for fast battery charging is becoming increasingly prominent. However, existing camera battery charging technologies mainly use fixed charging parameters, lacking the ability to dynamically perceive the actual state of the battery and intelligently adjust it, which can neither guarantee charging efficiency nor easily cause safety hazards such as battery overcharging or local overheating.

[0003] Traditional charging methods often ignore the battery's usage history and degree of wear, and are unable to develop personalized charging strategies for batteries in different states. At the same time, during the charging process, there is a lack of precise monitoring of temperature distribution and effective suppression of polarization effects, which not only affects the charging efficiency, but may also lead to a significant decrease in battery life, and cannot meet the strict requirements of professional camera equipment for battery performance and service life. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to achieve a balance between safe and efficient charging and battery life by real-time monitoring of battery status parameters and intelligent regulation of the charging process in combination with a neural network model.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a fast charging method for a camera smart battery, which includes collecting the current standby time and charge and discharge times of the camera battery, and generating a battery loss coefficient based on the standby time and the charge and discharge times; inputting the battery loss coefficient into a pre-trained neural network model, the neural network model establishes a mapping relationship between the battery loss coefficient and the safe charging power based on the historical operation data of the battery, and outputs a charging power upper limit value adapted to the current battery state through forward propagation calculation; collecting temperature distribution data of multiple preset detection points on the surface of the battery shell, matching the temperature distribution data with the charging power upper limit value, and generating a first round of dynamic charging power curve; in the execution stage of the first round of dynamic charging power curve, collecting voltage data at both ends of the battery to form a voltage response curve, and correcting the first round of dynamic charging power curve according to the trend characteristics of the voltage response curve; dividing the corrected charging power curve into multiple time periods according to preset time intervals, and charging the battery according to the power value corresponding to each time period until charging is completed.

[0008] As a preferred solution of the fast charging method for the camera intelligent battery described in the present invention, the step of generating a battery loss coefficient according to the standby time and the number of charge and discharge times includes: dividing the standby time into a first preset interval, a second preset interval and a third preset interval, the first preset interval corresponds to a first weight coefficient, the second preset interval corresponds to a second weight coefficient, and the third preset interval corresponds to a third weight coefficient; dividing the charge and discharge cycles into deep charging cycles and shallow charging cycles based on the previous charging cut-off voltages, the deep charging cycle corresponds to a deep charging attenuation coefficient, and the shallow charging cycle corresponds to a shallow charging attenuation coefficient; the battery loss coefficient is calculated by multiplying the interval weight coefficient of the standby time and the corresponding charging attenuation coefficient.

[0009] As a preferred solution of the fast charging method for the camera intelligent battery described in the present invention, the step of calculating and outputting the charging power upper limit value adapted to the current battery state through forward propagation includes: constructing a twin model structure based on the battery loss coefficient; judging the numerical range of the battery loss coefficient and selecting the corresponding sub-model; the battery loss coefficient is received by the input layer of the selected sub-model, and after the linear transformation of the hidden layer and the activation function processing, the charging power upper limit value is generated by the output layer.

[0010] As a preferred solution of the fast charging method for the camera intelligent battery described in the present invention, wherein: constructing a dual-model structure based on the battery loss coefficient includes: obtaining battery historical operation data, the battery historical operation data includes the battery loss coefficient and the corresponding charging parameters, the charging parameters include the charging current value, the charging time and the battery capacity value; dividing the battery historical operation data into a high-loss data subset and a low-loss data subset according to the loss threshold of the battery loss coefficient; constructing a network structure of the first sub-model and the second sub-model, the network structure includes an input layer, a hidden layer and an output layer; training the first sub-model with the high-loss data subset, the hidden layer of the first sub-model adopts the first activation function combination, and the loss function adopts the first loss function; training the second sub-model with the low-loss data subset, the hidden layer of the second sub-model adopts the second activation function combination, and the loss function adopts the second loss function; respectively recording the prediction errors of the first sub-model and the second sub-model on the corresponding loss data subsets, and stopping the training when the change in the prediction error of consecutive Kstop rounds of training is less than the preset error value.

[0011] As a preferred solution of the fast charging method of the camera intelligent battery described in the present invention, the step of generating a first-round dynamic charging power curve includes: setting a temperature sensor on the surface of the battery shell, the temperature sensor is set at the positive end, the negative end and the middle area of ​​the battery; collecting temperature data of each detection point through the temperature sensor, the temperature sensor includes detection units located at the positive end, the negative end and the middle area of ​​the battery; collecting temperature data of each detection point in real time through the temperature sensor, and calculating the temperature gradient value between adjacent detection points; constructing a temperature distribution matrix based on the temperature data; comparing the temperature distribution matrix with the standard temperature range to generate a temperature correction coefficient; multiplying the charging power upper limit value and the temperature correction coefficient to obtain the first-round dynamic charging power curve.

[0012] As a preferred solution of the fast charging method for the camera intelligent battery described in the present invention, the step of correcting the first round of dynamic charging power curve according to the trend characteristics of the voltage response curve includes: collecting the battery terminal voltage through a voltage sampling circuit, and the sampling frequency of the voltage sampling circuit is adjusted according to the voltage change rate; the battery terminal voltage data is plotted into a voltage response curve according to a time series, and a piecewise linear fitting method is used to calculate the slope change value of the voltage response curve in different time periods; based on the slope change value, the polarization characteristics of the battery are identified, a polarization state evaluation model is constructed, and a polarization degree coefficient is calculated; according to the polarization degree coefficient, the first round of dynamic charging power curve is adaptively corrected to generate a corrected charging power curve.

[0013] As a preferred solution of the fast charging method of the camera intelligent battery described in the present invention, the step of charging the battery according to the power value corresponding to each time period includes: obtaining a revised charging power curve; based on the revised charging power curve, dividing the charging process into a fast charging stage, a constant current stage and a trickle current stage; in the fast charging stage, outputting a first power value according to the revised charging power curve, wherein the first power value decreases in a step-like manner as the charging time passes; when the battery voltage reaches a preset conversion voltage, entering a constant current stage, and outputting a second power value lower than the first power value; in the trickle current stage, outputting a third power value for supplementary charging until the battery terminal voltage reaches the rated voltage or the charging current is lower than the preset cut-off current; in each time period, dynamically fine-tuning the charging power based on the battery terminal voltage and temperature data.

[0014] The beneficial effects of the present invention are as follows: the present invention introduces a neural network model to establish a mapping relationship between the battery loss coefficient and the safe charging power, thereby realizing intelligent adjustment of the charging parameters; adopts a twin model structure to train and optimize batteries with different loss degrees respectively, thereby improving the adaptability and accuracy of the charging strategy; through multi-point temperature monitoring and distribution analysis, accurate temperature management of the battery charging process is realized, and local overheating problems are effectively prevented; based on the voltage response curve, the battery polarization characteristics are analyzed, and the polarization effect in the charging process is effectively suppressed through adaptive power regulation; a segmented charging strategy is adopted to realize seamless switching among the three stages of fast charging, constant current and trickle current, thereby ensuring charging efficiency and extending the battery service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 The framework flow chart of the fast charging method of the camera smart battery.

[0017] Figure 2 The present invention is a flow chart for calculating the battery loss coefficient of the fast charging method of the camera intelligent battery.

[0018] Figure 3 Construct a flow chart for the battery loss twin model of the fast charging method of the camera smart battery. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, reference Figure 1~Figure 3, which is the first embodiment of the present invention, and this embodiment provides a method for fast charging of a camera intelligent battery, and the framework flow chart is as follows Figure 1 As shown, including

[0023] S1: Collect the current standby time and charge and discharge times of the camera battery, and generate a battery loss coefficient according to the standby time and charge and discharge times.

[0024] Specifically, the battery loss coefficient calculation flow chart is as follows Figure 2 As shown, specifically as follows: the standby time is divided into a first preset interval, a second preset interval and a third preset interval, the first preset interval corresponds to a first weight coefficient, the second preset interval corresponds to a second weight coefficient, and the third preset interval corresponds to a third weight coefficient.

[0025] Furthermore, the charge and discharge cycles are divided into deep charge cycles and shallow charge cycles based on the previous charge cut-off voltages. The deep charge cycle corresponds to a deep charge attenuation coefficient, and the shallow charge cycle corresponds to a shallow charge attenuation coefficient. The attenuation coefficient is obtained by experimental measurement based on the battery capacity attenuation curve. The battery loss coefficient is calculated by multiplying the interval weight coefficient of the standby time by the corresponding charging attenuation coefficient.

[0026] It should be noted that, in this embodiment, the first preset interval of the standby time corresponds to a short-term standby state (0 to 8 hours), the second preset interval corresponds to a medium-term standby state (8 to 16 hours), and the third preset interval corresponds to a long-term standby state (more than 16 hours). Correspondingly, the weight coefficient increases with the standby time, that is, the first weight coefficient is less than the second weight coefficient, and the second weight coefficient is less than the third weight coefficient. The incremental step size of the weight coefficient can be adjusted according to the actual application scenario, and it is recommended that the difference in weight coefficients between adjacent intervals is between 0.1 and 0.3. The charging cycle type is divided into deep charging and shallow charging based on a preset percentage of the battery rated voltage as a threshold (such as 95%), where the deep charging attenuation coefficient is greater than the shallow charging attenuation coefficient. The deep charging attenuation coefficient usually ranges from 0.8 to 1, and the shallow charging attenuation coefficient usually ranges from 0.5 to 0.7.

[0027] In addition, the battery loss coefficient is calculated as:

[0028]

[0029] Where D is the battery loss coefficient, W i is the interval weight coefficient corresponding to the standby time (i=1, 2, 3), N d is the number of deep charging times, N s is the number of shallow charges, K d is the deep charge attenuation coefficient, K sis the shallow charge attenuation coefficient, N is the total number of charges (N = N d + N s ).

[0030] It should be noted that the battery loss coefficient D in the formula ranges from 0 to 1, where 0 indicates that the battery performance has not been significantly degraded, and 1 indicates that the battery has reached the end of its service life.

[0031] Preferably, for the specific usage scenarios of camera batteries, an innovative battery loss coefficient calculation model based on the standby time partition weight and the charging depth attenuation coefficient is established. The battery loss coefficient is calculated by multiplying the interval weight coefficient and the charging attenuation coefficient, realizing an accurate quantitative evaluation of the battery status. This solution fully considers the loss characteristics of camera batteries in different standby states and different depths of charge and discharge cycles, making the calculation of the loss coefficient more in line with actual usage conditions.

[0032] S2: The battery loss coefficient is input into a pre-trained neural network model. The neural network model establishes a mapping relationship between the battery loss coefficient and the safe charging power based on the historical battery operation data, and outputs the charging power upper limit value that is adapted to the current battery status through forward propagation calculation.

[0033] Specifically, the following steps are included:

[0034] S2.1: Construct a twin model structure based on the numerical range of the battery loss coefficient.

[0035] Among them, the battery loss twin model construction flow chart is as follows Figure 3 As shown, the first sub-model is optimized for high loss coefficients, and the second sub-model is optimized for low loss coefficients. The specific design is as follows:

[0036] S2.1.1: Obtain historical battery operation data.

[0037] The battery historical operation data includes the battery loss coefficient and the corresponding charging parameters, and the charging parameters include charging current, charging time, battery capacity and charging temperature. The sampling period of historical operation data should not exceed 1 hour, and the data volume should be no less than 1,000 groups to ensure the effectiveness of model training.

[0038] S2.1.2: Divide the battery historical operation data into a high-loss data subset and a low-loss data subset according to the loss threshold of the battery loss coefficient.

[0039] The loss threshold of the battery loss coefficient is determined by analyzing the correlation between the battery capacity decay rate and the loss coefficient in the battery's historical operation data, and the loss coefficient value is specifically selected to make the inflection point position where the battery capacity decay rate increases significantly. Usually, the loss threshold is between 0.6 and 0.7, and the specific value can be obtained by analyzing historical data using the K-means clustering method.

[0040] S2.1.3: Construct the network structure of the first sub-model and the second sub-model.

[0041] Among them, the network structure includes an input layer, a hidden layer and an output layer, and the specific design is as follows: the input layer includes n neurons, receives an n-dimensional feature vector, and the input features include the battery loss coefficient, the current voltage value and the current temperature value; the hidden layer adopts a three-layer structure, and the number of neurons in each layer is 2n, 4n and 2n respectively; the output layer includes one neuron, which outputs the upper limit of the charging power. The first sub-model and the second sub-model use the same network structure, but use different activation function combinations and loss functions for targeted optimization. The value range of n is recommended to be 3 to 5, which can be adjusted according to the actual feature dimension. The Dropout layer (with a dropout rate of 0.2 to 0.3) can be optionally added to the network structure to prevent overfitting.

[0042] S2.1.4: Train the first sub-model using the high-loss data subset.

[0043] The first activation function combination is: the first hidden layer uses the ReLU function, the second hidden layer uses the Tanh function, and the third hidden layer uses the Sigmoid function. The first loss function uses the root mean square error function (RMSE) to reduce the power prediction deviation under high loss conditions. The initial value of the learning rate during training is set to 0.001, the batch size is recommended to be 32 or 64, and the upper limit of the training round is set to 1000 rounds.

[0044] S2.1.5: Train the second sub-model using the low-loss data subset.

[0045] Specifically, the hidden layer of the second sub-model uses the second activation function combination, and the loss function uses the second loss function. Among them, the second activation function combination is: the three hidden layers all use the ReLU function; the second loss function uses the mean absolute error function (MAE) to improve the prediction accuracy under low loss conditions. The training process uses the small batch gradient descent method, and the Adam optimizer is used to achieve adaptive adjustment of the learning rate. In order to improve the generalization ability of the model, the L2 regularization method can be used, and the regularization coefficient The value range is 0.0001 to 0.001.

[0046] S2.1.6: Record the prediction errors of the first sub-model and the second sub-model on the corresponding loss data subsets respectively. stop When the change in prediction error of each round of training is less than the preset error value, the training is stopped.

[0047] It should be noted that K stop To train the number of early stop rounds, the optimal value is selected in the interval [5, 20] through cross-validation method to prevent overfitting while ensuring model convergence. The preset error value is recommended to be set to 1% of the current prediction error. If the error change of consecutive Kstop rounds is less than this threshold, the early stop mechanism is triggered.

[0048] S2.1.7: Verify the trained first sub-model and the second sub-model.

[0049] Specifically, use the reserved validation data set to evaluate the performance of the two sub-models respectively, record the prediction error on the validation set, and complete the model validation when the validation error meets the preset accuracy requirement. If the validation error exceeds the preset range, return to S2.1.4 or S2.1.5 to perform parameter tuning and retraining on the corresponding sub-model. The validation set should account for 20% of the total data set, and the ratio of validation error to training error should not exceed 1.2, otherwise it indicates that the model may have an overfitting problem.

[0050] Preferably, the present invention adopts a twin model structure to optimize high and low loss states respectively, and realizes accurate prediction of charging power of batteries with different loss degrees through differentiated activation function combination and loss function design. This solution overcomes the technical problem that a single model is difficult to take into account the characteristics of high and low loss states at the same time, and improves the adaptability and accuracy of charging power prediction through model differentiation.

[0051] S2.2: Determine the numerical range of the battery loss coefficient and select the corresponding sub-model.

[0052] Specifically, the input battery loss coefficient is compared with the loss threshold, and the first sub-model is selected when the loss coefficient is greater than the loss threshold, and the second sub-model is selected when the loss coefficient is less than or equal to the loss threshold.

[0053] S2.3: After the input features are standardized and preprocessed, they are received by the input layer of the selected sub-model, and after the linear transformation and activation function processing of the hidden layer, the charging power upper limit value is generated by the output layer.

[0054] Specifically, the validity of the upper limit of the charging power output by the model is verified to ensure that it meets the preset safety range requirements. If it exceeds the safety range, the boundary value of the safety range is taken as the final output. The upper limit of the safety range should not exceed 80% of the rated charging power of the battery, and the lower limit should not be less than 20% of the rated charging power.

[0055] S3: Collect temperature distribution data of multiple preset detection points on the surface of the battery casing, match the temperature distribution data with the upper limit of the charging power, and generate the first round of dynamic charging power curve.

[0056] Specifically, the following steps are included:

[0057] S3.1: Multiple temperature sensors are evenly arranged along the circumference of the battery housing surface.

[0058] Specifically, the temperature sensor includes detection units located at the positive end, negative end and middle area of ​​the battery. In addition, the number of temperature detection points is not less than 6, and the distance between adjacent detection points is not greater than 1 / 6 of the circumference of the battery shell.

[0059] In addition, the temperature sensor uses an NTC thermistor with a temperature measurement accuracy of no less than ±0.5°C and a sampling frequency of no less than 1Hz. The sensor and the battery casing surface are fixed with thermally conductive silicone to ensure good thermal conductivity.

[0060] S3.2: The temperature data of each detection point is collected in real time through the temperature sensor, and the temperature gradient value between adjacent detection points is calculated.

[0061] The temperature gradient value The calculation formula is:

[0062]

[0063] Among them, T i and T i+1 are the temperature values ​​of two adjacent detection points respectively.

[0064] In addition, digital filtering is required when collecting temperature data. The 5-point median filtering method can be used to eliminate outliers, and the temperature difference of three consecutive samples at each detection point should not exceed 2°C, otherwise it will be judged as a sensor abnormality.

[0065] S3.3: Construct the battery surface temperature distribution matrix based on the temperature data.

[0066] Specifically, the temperature distribution matrix reflects the temperature uniformity of each area of ​​the battery. The construction process is as follows: the temperature data of each detection point is arranged into an n×1 column vector T according to the spatial position; the temperature gradient value of each detection point and its adjacent points is calculated to form a gradient matrix G; the temperature vector T and the gradient matrix G are combined to construct an n×n temperature distribution matrix M, where Represents the temperature change characteristics from the i-th detection point to the j-th detection point. The temperature distribution matrix M is constructed using the weighted average method. The weight coefficient is inversely proportional to the distance between the detection points. The closer the distance, the greater the weight. The diagonal elements are set to 1, and the value range of the non-diagonal elements is [0, 1].

[0067] S3.4: Compare the temperature distribution matrix with the preset standard temperature range to generate a temperature correction coefficient.

[0068] The temperature correction coefficient decreases as the degree of temperature distribution non-uniformity increases, and the generation process of the temperature correction coefficient is as follows: calculate the eigenvalue of the temperature distribution matrix M and obtain its maximum eigenvalue and the minimum eigenvalue ; Define the temperature distribution unevenness index ; Temperature correction factor The calculation formula is: , where k is the attenuation coefficient, determined by experimental calibration, It is an index of temperature distribution unevenness. The range of the attenuation coefficient k is [0.5, 2], and the k value can be dynamically adjusted according to the battery type.

[0069] It should be noted that the standard temperature range is defined as ,in is the minimum safe operating temperature, The maximum safe working temperature is generally in the range of [0°C, 45°C]. When the temperature of any detection point exceeds the standard temperature range, the temperature correction coefficient Directly take 0 and stop charging. The temperature gradient threshold of adjacent detection points is set to 5°C. When the temperature gradient value exceeds this threshold, the temperature balance protection is triggered. In addition, the triggering of the temperature balance protection not only considers the temperature gradient, but also the temperature change rate. When the temperature change rate of any detection point exceeds 2°C / min, the protection is also triggered.

[0070] S3.5: Multiply the upper limit value of the charging power by the temperature correction coefficient to obtain a charging power curve after temperature compensation.

[0071] It should be noted that the charging power after temperature compensation satisfies the step-by-step change. The amplitude of each power adjustment should not exceed 20% of the current power, and the time interval between adjacent adjustments should not be less than 30s to avoid power mutation causing impact on the battery.

[0072] S3.6: Use the charging power curve after temperature compensation as the first round of dynamic charging power curve.

[0073] It should be noted that the first round of dynamic charging power curve decreases as the unevenness of battery temperature distribution increases.

[0074] In particular, the temperature distribution is continuously monitored during the charging process, and when the temperature distribution changes significantly ( The rate of change of exceeds the rate of change threshold), updates the temperature correction coefficient in real time, and dynamically adjusts the charging power curve. The change rate threshold is set to 0.1 / min. When the deviation between the updated charging power and the current power exceeds 30%, a segmented transition method is adopted to gradually adjust to the target power value, and the transition time is not less than 1 minute.

[0075] S4: During the charging process, a voltage response curve of the battery is collected, and the first round charging power curve is corrected according to the voltage response curve.

[0076] Specifically, the following steps are included:

[0077] S4.1: When the charger is charging the battery according to the first round of dynamic charging power curve, the battery terminal voltage is collected through the voltage sampling circuit.

[0078] Among them, the sampling frequency of the voltage sampling circuit is adaptively adjusted according to the voltage change rate. It should be noted that when the voltage change rate is greater than the preset voltage change rate threshold, the sampling frequency is increased to the high-frequency sampling frequency, otherwise the reference sampling frequency is used. Among them, the high-frequency sampling frequency is set to 20Hz, and the reference sampling frequency is set to 2Hz. At the same time, current sampling and voltage sampling are synchronized to accurately calculate the polarization characteristics.

[0079] S4.2: Plot the collected battery terminal voltage data into a voltage response curve in time series, and use a piecewise linear fitting method to calculate the slope change value of the voltage response curve in different time periods.

[0080] Specifically, the charging process is divided into n equal time intervals (n ≥ 10), and the least squares method is used for linear fitting in each time interval to obtain the slope value of the time period, and the slope change of adjacent time intervals is calculated. The division of time intervals needs to consider the battery capacity and charging rate, and the general range is 30 to 60 seconds. In addition, the number of data points in each time interval is not less than 50, and the goodness of fit R 2 If the residual error is greater than 0.95, the time interval is re-divided. Data points whose residual error exceeds 3 times the standard deviation of the fitting result are marked as outliers and removed.

[0081] S4.3: Identify the concentration polarization and electrochemical polarization characteristics of the battery based on the slope change value and charging current data, and construct a polarization state evaluation model.

[0082] Specifically, the polarization process is divided into a concentration polarization stage and an electrochemical polarization stage according to the slope change characteristics: wherein the concentration polarization characteristics are characterized by a fast rising section of the voltage response curve, and the corresponding slope change is greater than a first preset threshold value ΔK 1 (generally 0.5 V / min); the electrochemical polarization characteristics are characterized by the slow rise of the voltage response curve, and the corresponding slope change is less than the second preset threshold ΔK 2 (Generally take 0.1V / min) interval.

[0083] Furthermore, the design of the polarization state evaluation model is based on the physical and chemical mechanism of the battery charging process, and a composite function is used to comprehensively characterize the polarization characteristics of different types. The exponential form is used because the concentration polarization increases exponentially with the current density, reflecting the rapid polarization process caused by the limited diffusion of ions in the electrolyte; the electrochemical polarization function The logarithmic form corresponds to the asymptotic characteristics of charge transfer on the electrode surface, reflecting the logarithmic relationship between activation polarization and overpotential; the time correction function The exponential decay form is used to characterize the time accumulation effect of the polarization process, reflecting the characteristic that the degree of polarization will gradually stabilize over time. The three functions are combined through weight coefficients, which not only takes into account the relative importance of various polarization mechanisms, but also ensures the applicability and robustness of the model. The ratio of weight coefficients is mainly based on the dominant role of concentration polarization in the fast charging process, while taking into account the influence of electrochemical polarization and time accumulation effect. The specific formula is as follows:

[0084]

[0085] in, It is the concentration polarization function, which is proportional to the slope change of the fast rising section of the voltage response curve; It is the electrochemical polarization function, which is proportional to the slope change of the slowly rising section of the voltage response curve; is the time correction function; , and is the weight coefficient, which is determined by experimental calibration and satisfies ; is the total polarization state value.

[0086] In this embodiment, , , ,parameter Obtained through fitting of a large amount of experimental data. Recommended values ​​of weight coefficient: , , .

[0087] S4.4: Calculate the polarization degree coefficient based on the polarization state evaluation model.

[0088] Specifically, the polarization coefficient The calculation formula is:

[0089]

[0090] in, is the polarization attenuation factor, which is obtained through the battery polarization characteristic experimental calibration, and the general value range is [0.1, 0.5]. is the total polarization state value.

[0091] It should be noted that the polarization coefficient The value range of is [0, 1]. The more serious the polarization, The smaller the value, the The value is less than the safety threshold (0.3 is recommended), the polarization protection mechanism is triggered and the charging power is forced to be reduced.

[0092] S4.5: Adaptively correct the first round of dynamic charging power curve according to the polarization degree coefficient to generate a corrected charging power curve.

[0093] S5: Divide the corrected charging power curve into a plurality of time periods, and charge the battery according to the power value corresponding to each time period until charging is completed.

[0094] Specifically, a corrected charging power curve is obtained; based on the corrected charging power curve, the charging process is divided into a fast charging stage, a constant current stage, and a trickle current stage. The specific division method is as follows: The stage division of the charging process is based on the output value of the battery state of charge (SOC) and polarization state evaluation model Comprehensive determination: and It is the fast charging stage; when or It is the constant current stage; when or The SOC is calculated by the ampere-hour integration method combined with the open circuit voltage correction, and the integration error should be controlled within 1%.

[0095] In the fast charging stage, the charger outputs a first power value according to the revised charging power curve, and the first power value decreases in steps with the charging time, where the power reduction step is 10%~15% of the initial charging power, and the duration of each step is 10~15 minutes; when the battery voltage reaches the preset conversion voltage, it enters the constant current stage, and the charger outputs a second power value according to the revised charging power curve, and the second power value is lower than the first power value; in the trickle charging stage, the charger outputs a third power value for supplementary charging, continuously detects the battery terminal voltage value, and stops charging when the battery terminal voltage value reaches the rated voltage or the charging current is detected to be lower than the preset cut-off current.

[0096] Furthermore, in each time period, the charging power is dynamically fine-tuned based on the battery terminal voltage and temperature data to ensure the charging process is safe and reliable. The dynamic fine-tuning of the charging power adopts a fuzzy control strategy, which uses the voltage change rate as the and temperature change rate As input variable, output power adjustment amount. The specific adjustment rule is: or When the power is reduced by 15%~20%; when or When the power is reduced by 5%~10%; when and When the current power is maintained unchanged.

[0097] In addition, the fuzzy controller adopts the Mamdani reasoning method, with a control period of 1s and a minimum adjustment step of 1%. At the same time, a self-recovery mechanism is set to allow the power to be gradually restored when the state remains stable.

[0098] In summary, the present invention introduces a neural network model to establish a mapping relationship between the battery loss coefficient and the safe charging power, thereby realizing intelligent adjustment of the charging parameters; adopts a twin model structure to train and optimize batteries with different loss degrees respectively, thereby improving the adaptability and accuracy of the charging strategy; through multi-point temperature monitoring and distribution analysis, accurate temperature management of the battery charging process is achieved, effectively preventing local overheating problems; based on the voltage response curve, the battery polarization characteristics are analyzed, and the polarization effect in the charging process is effectively suppressed through adaptive power regulation; a segmented charging strategy is adopted to realize seamless switching among the three stages of fast charging, constant current and trickle current, which not only ensures the charging efficiency but also prolongs the battery life.

[0099] Example 2, reference Figure 1~Figure 3 , which is the second embodiment of the present invention, and this embodiment provides a fast charging method for a camera intelligent battery. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0100] To verify the effectiveness of the fast charging method, the research team selected 5 sets of professional camera lithium batteries with a capacity of 4800mAh for comparative experiments. The experimental environment temperature was controlled at 25±2℃ and the relative humidity was 50±5%. The test equipment included a programmable DC power supply (Keysight N6705C), a high-precision battery test system (Arbin BT2000) and a multi-channel temperature acquisition instrument (Agilent 34972A).

[0101] In the battery loss coefficient calculation stage, the standby time data of the battery was first collected, and the weight coefficients of the three intervals of 0-8 hours, 8-16 hours, and more than 16 hours were set to 0.3, 0.5, and 0.7 respectively. The attenuation coefficients of deep charging and shallow charging were calibrated by experiments to be 0.82 and 0.65 respectively. The average loss coefficient of the five groups of batteries measured by the experiment was 0.48, indicating that the battery was in normal use.

[0102] The neural network model was trained using 1,000 sets of historical operating data, with the loss threshold determined to be 0.65 through K-means clustering. The input layer uses three neurons, corresponding to the loss coefficient, voltage value, and temperature value respectively. The number of neurons in the hidden layer is configured as 6-12-6, the learning rate is set to 0.001, and the batch size is 32. After 780 rounds of training, the root mean square error of the model on the validation set dropped to 3.1%, meeting the preset accuracy requirements.

[0103] The temperature distribution detection uses 8 NTC thermistors (accuracy ±0.3℃), which are evenly arranged along the battery shell. During the experiment, the maximum temperature difference on the battery surface was controlled within 5.2℃, and the temperature gradient value did not exceed the preset threshold of 5℃. The peak value of the temperature distribution unevenness index was 0.18, and the corresponding temperature correction coefficient was 0.89, and the charging power was reduced by 11% accordingly.

[0104] In order to comprehensively evaluate the performance advantages of the method of the present invention, the research team conducted a detailed comparison between it and the traditional CC-CV method and the constant power fast charging method. The comparison results of the main performance indicators are shown in Table 1.

[0105] Table 1 Comparison of main performance indicators of the solution of the present invention and the prior art

[0106]

[0107] The results of the comparative experiments show that compared with the traditional CC-CV method and the constant power fast charging method, the intelligent fast charging method proposed in the present invention has certain advantages in terms of charging time, temperature control, cycle life and charging efficiency. In particular, in terms of polarization suppression effect, the method of the present invention reduces the polarization voltage to 112mV, which is 32.1% lower than the traditional method, thanks to the accurate polarization state evaluation and adaptive power regulation mechanism.

[0108] In the voltage response characteristic analysis, the voltage sampling frequency is increased to 20Hz when the voltage changes drastically, and maintained at 2Hz under normal conditions. The slope change value obtained by piecewise linear fitting fluctuates in the range of 0.08-0.52V / min. The concentration polarization and electrochemical polarization characteristics identified based on this are consistent with the theoretical analysis. The weight coefficients of the polarization state evaluation model were optimized and determined to be: α=0.45, β=0.35, γ=0.2. The calculated polarization degree coefficient always remains above 0.42, and the polarization protection mechanism is not triggered.

[0109] The final charging process is divided into three stages: the fast charging stage (0-55 minutes) starts at 55W and then decreases step by step, with each stage lasting 12 minutes; the constant current stage (55-80 minutes) is maintained at 28W; the trickle stage (80-92 minutes) drops to 12W until charging is complete. The battery temperature reached a maximum of 42.5°C throughout the process, which did not exceed the safety threshold. The capacity test showed that the charging time to 100% SOC was 92±3 minutes, the charging efficiency reached 91%, and the capacity retention rate remained above 94.2% after 100 cycle tests.

[0110] This fast charging solution has demonstrated good performance and stability in actual applications. It shortens the charging time while ensuring safety and extends the battery life to a certain extent, providing new ideas for the development of fast charging technology for professional camera lithium batteries.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A fast charging method for a camera smart battery, characterized in that: include, Collect the current standby time and charge and discharge times of the camera battery, and generate a battery loss coefficient according to the standby time and the charge and discharge times; The battery loss coefficient is input into a pre-trained neural network model, the neural network model establishes a mapping relationship between the battery loss coefficient and the safe charging power based on the historical battery operation data, and outputs a charging power upper limit value adapted to the current battery state through forward propagation calculation; Collecting temperature distribution data of multiple preset detection points on the surface of the battery housing, matching the temperature distribution data with the upper limit of the charging power, and generating a first round of dynamic charging power curve; In the execution phase of the first round of dynamic charging power curve, voltage data at both ends of the battery are collected to form a voltage response curve, and the first round of dynamic charging power curve is corrected according to the trend characteristics of the voltage response curve; Dividing the corrected charging power curve into a plurality of time periods according to preset time intervals, and charging the battery according to the power value corresponding to each time period until charging is completed; The step of generating a first round of dynamic charging power curve comprises: A temperature sensor is arranged on the surface of the battery housing, wherein the temperature sensor is arranged at the positive terminal, the negative terminal and the middle area of ​​the battery; The temperature data of each detection point is collected by a temperature sensor, wherein the temperature sensor includes detection units located at the positive end, the negative end and the middle area of ​​the battery; The temperature sensor collects the temperature data of each detection point in real time and calculates the temperature gradient value between adjacent detection points; Constructing a temperature distribution matrix based on the temperature data; Comparing the temperature distribution matrix with a standard temperature range to generate a temperature correction coefficient; The charging power upper limit value is multiplied by the temperature correction coefficient to obtain a first-round dynamic charging power curve; The step of correcting the first-round dynamic charging power curve according to the trend characteristics of the voltage response curve includes: The battery terminal voltage is collected by a voltage sampling circuit, wherein the sampling frequency of the voltage sampling circuit is adjusted according to the voltage change rate; The battery terminal voltage data is plotted into a voltage response curve in time series, and a piecewise linear fitting method is used to calculate the slope change value of the voltage response curve in different time periods; Identify the polarization characteristics of the battery based on the slope change value, construct a polarization state evaluation model, and calculate the polarization degree coefficient; The first round of dynamic charging power curve is adaptively corrected according to the polarization degree coefficient to generate a corrected charging power curve.

2. The method for fast charging a camera smart battery as claimed in claim 1, characterized in that: The step of generating a battery loss coefficient according to the standby time and the number of charge and discharge times comprises: Dividing the standby time into a first preset interval, a second preset interval and a third preset interval, the first preset interval corresponds to a first weight coefficient, the second preset interval corresponds to a second weight coefficient, and the third preset interval corresponds to a third weight coefficient; The charge-discharge cycle is divided into a deep charge cycle and a shallow charge cycle based on the previous charge cut-off voltages, wherein the deep charge cycle corresponds to a deep charge attenuation coefficient, and the shallow charge cycle corresponds to a shallow charge attenuation coefficient; The battery loss coefficient is calculated by multiplying the interval weight coefficient of the standby time by the corresponding charging attenuation coefficient.

3. The fast charging method for a camera smart battery as claimed in claim 1, characterized in that: The step of calculating and outputting the upper limit value of the charging power adapted to the current battery state through forward propagation comprises: Construct a twin model structure based on the battery loss coefficient; Determine the numerical range of the battery loss coefficient and select the corresponding sub-model; The battery loss coefficient is received by the input layer of the selected sub-model, and after being processed by the linear transformation and activation function of the hidden layer, the charging power upper limit value is generated by the output layer.

4. The fast charging method for a camera smart battery as claimed in claim 3, characterized in that: The twin model structure constructed based on the battery loss coefficient includes: Acquire battery historical operation data, wherein the battery historical operation data includes a battery loss coefficient and corresponding charging parameters, wherein the charging parameters include a charging current value, a charging time, and a battery capacity value; Dividing the battery historical operation data into a high-loss data subset and a low-loss data subset according to a loss threshold of the battery loss coefficient; Constructing a network structure of the first sub-model and the second sub-model, wherein the network structure includes an input layer, a hidden layer, and an output layer; The first sub-model is trained using the high-loss data subset, wherein the hidden layer of the first sub-model adopts a first activation function combination, and the loss function adopts a first loss function; The second sub-model is trained using the low-loss data subset, wherein the hidden layer of the second sub-model adopts a second activation function combination, and the loss function adopts a second loss function; The prediction errors of the first sub-model and the second sub-model on the corresponding loss data subset are recorded respectively, and the training is stopped when the changes in the prediction errors of consecutive Kstop rounds of training are all less than the preset error value.

5. The method for fast charging a camera smart battery as claimed in claim 1, characterized in that: The step of charging the battery according to the power value corresponding to each time period includes: Obtain a corrected charging power curve; Based on the modified charging power curve, the charging process is divided into a fast charging stage, a constant current stage and a trickle current stage; In the fast charging stage, a first power value is output according to the modified charging power curve, wherein the first power value decreases in a step-like manner as the charging time passes; When the battery voltage reaches the preset conversion voltage, it enters the constant current stage and outputs a second power value lower than the first power value; In the trickle stage, the third power value is output for supplementary charging until the battery terminal voltage reaches the rated voltage or the charging current is lower than the preset cut-off current; During each time period, the charging power is dynamically fine-tuned based on the battery terminal voltage and temperature data.

Images