BMS-based battery health state evaluation method and system

Abstract

The invention provides a battery health state assessment method and system based on a BMS, and the method comprises the steps: presetting different voltage sampling ranges and resolutions for different load current intervals, judging the current interval to which the load current belongs when the load current changes, and dynamically switching to the corresponding range and resolution; when the current changes violently, the voltage sampling frequency is improved, more data points are obtained, and when the current changes gently, the sampling frequency is reduced, and data redundancy is reduced; filtering and fusing the internal resistance estimated value by adopting a Kalman filtering algorithm, dynamically tracking the change trend of the internal resistance, and adaptively adjusting the parameters of a filter according to the working state and the aging degree of the battery; in combination with working condition parameters including the temperature, the current and the SOC of the battery, a multiple regression model of the internal resistance and the working condition parameters is established, the multiple regression model is trained and optimized through a machine learning method, and online self-adaptive correction of the internal resistance is achieved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a battery health status assessment method and system based on BMS. Background Art

[0002] A key technical difficulty in battery health status assessment is how to accurately obtain internal resistance information. Internal resistance is an important parameter for evaluating battery health status, and its change is closely related to the battery's capacity decay and power decay. However, the measurement of internal resistance is very sensitive to the load current. When the load current is large, the operating voltage of the battery will have a significant voltage drop, which will directly affect the accuracy of voltage sampling and thus the accuracy of internal resistance calculation. When the load current is small, the battery voltage drop becomes less obvious, and the voltage response signal is weak, which is not conducive to the measurement of internal resistance. How to ensure the accuracy of voltage sampling and accurately obtain internal resistance information when the load current changes greatly is a technical problem that needs to be solved urgently. The core of this problem is that the voltage sampling circuit and the internal resistance calculation algorithm cannot adapt to the dynamic changes of the load current. The ideal voltage sampling circuit should be able to adjust the range and resolution in real time according to the load conditions to meet different measurement requirements. The internal resistance calculation algorithm needs to be able to comprehensively analyze the data of multiple sampling points, and minimize the impact of current changes on the accuracy of internal resistance estimation through certain mathematical models and calculation methods. Summary of the invention

[0003] The present invention provides a battery health status assessment method based on BMS, which mainly includes:

[0004] Obtain multiple sets of voltage sampling data of the battery under different load currents, calculate the corresponding internal resistance estimation value based on the voltage sampling data, build a mathematical model between internal resistance, current and voltage, and determine the model parameters through curve fitting method; preset different voltage sampling ranges and resolutions for different load current intervals, and when the load current changes, determine the current interval to which it belongs, and dynamically switch to the corresponding range and resolution; when the current changes drastically, increase the voltage sampling frequency to obtain more data points, and when the current changes slowly, reduce the sampling frequency to reduce data redundancy; use the Kalman filter algorithm to filter and fuse the internal resistance estimation value, dynamically track the change trend of the internal resistance, and adaptively adjust the filter parameters according to the working state and aging degree of the battery; combine the operating parameters including battery temperature, current, and SOC to establish a multivariate regression model of internal resistance and operating parameters, train and optimize the multivariate regression model through machine learning methods, and realize online adaptive correction of internal resistance; perform statistical analysis on the calculated internal resistance value, extract characteristic parameters including internal resistance mean, variance, and change rate, and combine the empirical rules in the expert knowledge base to comprehensively judge the health status of the battery, and generate a health assessment report and maintenance recommendations.

[0005] The present invention provides a battery health status assessment system based on BMS, which mainly includes: a data acquisition and modeling module, which is used to obtain voltage sampling data and construct a mathematical model of internal resistance, current and voltage; a dynamic sampling control module, which is used to dynamically adjust the voltage sampling range, resolution and frequency according to the load current change; a data filtering and fusion module, which is used to filter and dynamically track the internal resistance estimation value using the Kalman filter algorithm; an adaptive correction module, which is used to establish a multivariate regression model of internal resistance and operating condition parameters and realize online adaptive correction; a health status assessment module, which is used to analyze the internal resistance characteristic parameters and comprehensively judge the battery health status in combination with expert knowledge. The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0006] The present invention discloses a battery health status assessment method based on BMS, which constructs a mathematical model of internal resistance, current and voltage by acquiring voltage sampling data under different load currents. In response to changes in load current, the voltage sampling range, resolution and frequency are dynamically adjusted to obtain more accurate data. The Kalman filter algorithm is used to filter and fuse the internal resistance estimation value, and a multivariate regression model is established in combination with operating parameters such as battery temperature, current, SOC, etc. to achieve online adaptive correction of the internal resistance. The present invention also performs statistical analysis on the calculated internal resistance value, extracts characteristic parameters and combines with an expert knowledge base to comprehensively evaluate the battery health status. The method can accurately estimate the battery internal resistance, dynamically track its changing trend, and provide health assessment reports and maintenance recommendations, effectively improving the performance and reliability of the battery management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 The present invention is a flow chart of a method for evaluating the battery health status based on BMS.

[0008] Figure 2 A schematic diagram of a battery health status assessment method and system based on BMS of the present invention.

[0009] Figure 3 This is another schematic diagram of a battery health status assessment method and system based on BMS of the present invention.

[0010] Figure 4 The figure is a structural schematic diagram of a battery health status assessment system based on BMS of the present invention. DETAILED DESCRIPTION

[0011] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0012] like Figure 1-4 In this embodiment, a battery health status assessment system based on BMS may specifically include:

[0013] Step S101, obtaining multiple groups of voltage sampling data of the battery under different load currents, calculating the corresponding internal resistance estimation value according to the voltage sampling data, constructing a mathematical model between the internal resistance, current and voltage, and determining the model parameters by a curve fitting method.

[0014] A1. Obtain the load current variation range of the battery, and design at least two different current values ​​according to the load current variation range to conduct a discharge test; A2. For each current value, collect the voltage data of the battery at the current value to obtain multiple sets of current-voltage data pairs; A3. According to Ohm's law, use each set of current-voltage data pairs to calculate the estimated internal resistance of the battery under the working condition to obtain multiple internal resistance estimates; A4. According to multiple internal resistance estimates, multiple current values ​​and corresponding voltage data, establish a polynomial function relationship model between internal resistance, current and voltage; A5. Use the least squares method to perform parameter identification on the polynomial function relationship model to obtain a fitted internal resistance estimation model; A6. Obtain a preset data set, which includes multiple sets of current -voltage data pair; A7, using the leave-one-out cross-validation method, divide the data set into a training set and a test set; A8, use the training set to train the internal resistance estimation model, and use the test set to test the internal resistance estimation model to obtain the root mean square error and the determination coefficient; if the root mean square error is less than the first preset threshold, and the determination coefficient is greater than the second preset threshold, then embed the internal resistance estimation model into the battery management system; if the root mean square error is greater than or equal to the first preset threshold, or the determination coefficient is less than or equal to the second preset threshold, then return to step A5; A9, during operation, collect the current and voltage data of the battery in real time, substitute them into the internal resistance estimation model, and calculate the battery internal resistance under the current working conditions; A10, dynamically adjust the battery charging and discharging strategy according to the changes in the battery internal resistance.

[0015] Specifically, the load current variation range of the battery is obtained to cover various working conditions that the battery may encounter in actual use. For example, the current of electric vehicles under different working conditions such as starting, accelerating, driving at a constant speed, and braking varies greatly, and the current of mobile phones in standby, talking, playing games, etc. is also different. After determining the current variation range, it is necessary to select several representative current values ​​for testing. Assume that the measured current variation range is 1A to 5A. Then, four current values ​​of 1A, 2A, 3A, and 4A can be selected. For each current value, the voltage data of the battery at the current value is collected. For example, at a current of 1A, the voltage is measured to be 8V; at a current of 2A, the voltage is measured to be 7V; at a current of 3A, the voltage is measured to be 6V; at a current of 4A, the voltage is measured to be 5V. In this way, four sets of current-voltage data pairs are obtained: (1A, 8V), (2A, 7V), (3A, 6V), (4A, 5V). According to Ohm's law, the internal resistance of a battery is equal to the voltage drop of the battery divided by the corresponding current change. Taking the two sets of data (1A, 8V) and (2A, 7V) as examples, the voltage drop is 8V-7V=1V, and the current change is 2A-1A=1A, so the estimated internal resistance of the battery is 1V / 1A=1Ω. Similarly, the estimated internal resistance under other working conditions can be calculated. Using multiple sets of current-voltage data pairs and the calculated estimated internal resistance, a polynomial function relationship model between internal resistance, current and voltage can be established. The polynomial function is chosen because it has good fitting ability and can better describe the nonlinear relationship between internal resistance and current and voltage. For example, a quadratic polynomial model can be selected: R=a+bI+cV+dI^2+eV^2+fIV, where R is the internal resistance, I is the current, V is the voltage, and a, b, c, d, e, and f are unknown coefficients. The least squares method is used to identify the parameters of the polynomial function relationship model in order to find the optimal coefficients so that the model can better fit the actual data. The principle of the least squares method is to minimize the sum of squares of the errors between the model prediction value and the actual value. Through the least squares method, the values ​​of the coefficients a, b, c, d, e, and f in the model can be determined. The preset data set is obtained to verify the established internal resistance estimation model. The data set should contain multiple sets of current-voltage data pairs, which should cover the working state of the battery under different working conditions as much as possible. The leave-one-out cross-validation method is used to divide the data set into a training set and a test set. For example, if there are 10 sets of data, 9 sets of data are taken as training sets each time, and the remaining 1 set of data is taken as a test set, and a total of 10 training and testing are performed. The internal resistance estimation model is trained using the training set, that is, the coefficients in the model are optimized using the data of the training set. The internal resistance estimation model is tested using the test set, and the root mean square error and the coefficient of determination are calculated to evaluate the performance of the model. The root mean square error reflects the deviation between the model prediction value and the actual value, and the coefficient of determination reflects the goodness of fit of the model.The first preset threshold and the second preset threshold are set to determine whether the model meets the requirements. If the root mean square error is less than the first preset threshold and the determination coefficient is greater than the second preset threshold, it is considered that the accuracy and goodness of fit of the model have met the requirements, and the model can be embedded in the battery management system. Otherwise, it is necessary to return to re-execute the parameter identification step, adjust the model parameters or select a more suitable model structure. During the operation of the battery management system, the current and voltage data of the battery are collected in real time, substituted into the internal resistance estimation model, and the battery internal resistance under the current working conditions is calculated. According to the change of the battery internal resistance, the battery charging and discharging strategy can be dynamically adjusted. For example, when the internal resistance increases, it indicates that the health state of the battery is reduced, and the charging current can be reduced or charging can be stopped to protect the battery. The advantage of doing this is that it can extend the service life of the battery and improve the safety of the battery. The internal resistance estimation model established through the above steps can accurately estimate the internal resistance of the battery in real time, provide important parameter basis for the battery management system, thereby optimizing the battery charging and discharging strategy and improving the performance and life of the battery.

[0016] Step S102, for different load current intervals, different voltage sampling ranges and resolutions are preset. When the load current changes, the current interval to which it belongs is determined, and the corresponding range and resolution are dynamically switched.

[0017] Get the real-time value of the load current sensor, compare it with multiple preset current interval thresholds, and determine the range to which the load current belongs; if the load current is within a preset current interval, find the voltage sampling range and ADC resolution parameters corresponding to the interval from the pre-established current interval-sampling parameter mapping table; send the found voltage sampling range and ADC resolution parameters to the voltage sampling module through the I2C bus interface, and dynamically adjust the amplification factor and ADC conversion bit number of its sampling circuit; during the voltage sampling process, continuously monitor the value change of the load current sensor at a preset time interval to determine whether the load current crosses the boundary of the current interval; if the load current crosses the boundary of the current interval, re-determine the interval to which the load current belongs, and update the parameter configuration of the voltage sampling module; collect voltage according to the preset time window The sampling data is divided into multiple continuous sub-intervals according to the changes of the current interval and sampling parameters in each time window; the voltage sampling data in each sub-interval is denoised and filtered, and the sampling value is linearly scaled according to the voltage sampling range and ADC resolution parameters of the sub-interval to obtain a unified physical voltage value; the average power and cumulative energy consumption in each sub-interval are calculated according to the average current value of the sub-interval and the corrected voltage sampling sequence; the average power and cumulative energy consumption of each sub-interval are accumulated to obtain the total power and total energy consumption of the load in the entire time window; the total power is compared with the preset overload threshold, and the total energy consumption is compared with the preset over-limit threshold to determine whether the load operation status is abnormal; if power overload or energy consumption exceeds the limit, an alarm signal is triggered and the preset overload protection control logic is started.

[0018] Specifically, in modern battery management systems, real-time monitoring of load current and dynamically adjusting voltage sampling parameters based on its changes is a key technology. For example, suppose a battery management system is designed to monitor a lithium battery that has different current outputs under different working conditions. The system first obtains the real-time current value through the load current sensor. If the current value read by the current sensor is 3 amps, and the system preset current intervals are 0-2 amps, 2-4 amps, and 4-6 amps, then 3 amps belongs to the second current interval. Based on this current interval, the system will look up the corresponding voltage sampling range and ADC resolution from a pre-set mapping table. Assume that for the 2-4 amp interval, the optimal voltage sampling range is 0-10 volts, and the ADC resolution is set to 12 bits. These parameters are then sent to the voltage sampling module via the I2C bus to ensure that the sampling accuracy matches the current load. During the voltage sampling process, the system detects the current value every 100 milliseconds to monitor whether there are any changes across intervals. If the current suddenly increases to 5 amps, the system will immediately detect this change and readjust the parameters of the voltage sampling module to adapt to the new current range of 4-6 amps. The corresponding voltage sampling range may be adjusted to 0-12 volts, and the ADC resolution is increased to 14 bits. When collecting voltage data, the system will record the voltage value in each window with a time window of 200 milliseconds. These data will be divided into different sub-intervals according to the change of current, and the current and sampling parameters in each sub-interval will be consistent. For the data of each sub-interval, the system will perform denoising and filtering, and then perform linear scale correction according to the sampling parameters to ensure the accuracy of the data. Using the corrected voltage and real-time current data, the system can calculate the average power and cumulative energy consumption of each sub-interval. For example, if the average voltage is 8 volts and the average current is 3 amps in a 200 millisecond time window, the average power is 24 watts. If this state lasts for 1 second, the cumulative energy consumption is 24 joules. The system accumulates the average power and cumulative energy consumption of all sub-intervals to obtain the total power and total energy consumption during the entire monitoring period. These data are compared with the system's preset overload and energy consumption thresholds to determine whether there are any abnormal conditions. If the total power exceeds the preset overload threshold, the system triggers an alarm signal and starts the overload protection control logic, such as reducing the battery's output power or suspending the battery's use, to protect the safety of the battery and the device. In this way, the battery management system can respond to current changes in real time, optimize the voltage sampling strategy, and effectively monitor the battery's energy consumption and output status, thereby improving the battery's use efficiency and extending its life.

[0019] Step S103, when the current changes dramatically, the voltage sampling frequency is increased to obtain more data points; when the current changes slowly, the sampling frequency is reduced to reduce data redundancy.

[0020] The current data of the current sensor is obtained, and the current data is subjected to sliding average filtering to obtain a smooth current curve; the change rate between two adjacent current data points is calculated. If the current change rate exceeds a preset threshold, the current is judged to be drastic; if the current change rate is lower than the preset threshold, the current is judged to be gentle; if the current change is judged to be drastic, the voltage sampling frequency is increased to a preset maximum value; if the current change is judged to be gentle, the voltage sampling frequency is reduced to a preset minimum value; the voltage sampling data of the voltage sensor is obtained through the ADC module; the voltage sampling data is filtered using the Kalman filter algorithm to obtain a smooth voltage curve. The Kalman filter estimates the optimal value of the voltage by recursion and removes high-frequency noise; the smoothed voltage curve is subjected to first-order difference to obtain a voltage change rate curve; the voltage change rate at the current moment is calculated, and compared with the preset voltage change rate threshold, and the voltage sampling frequency at the next moment is dynamically adjusted; if the voltage change rate is greater than the preset threshold, the sampling frequency is increased; if the voltage change rate is less than the preset threshold, the sampling frequency is reduced.

[0021] Specifically, the raw current data obtained by the current sensor often contains noise and fluctuations. In order to obtain a more stable current value, the current data needs to be processed by sliding average filtering. The principle of sliding average filtering is to average the current value at the current moment with the current values ​​at several previous moments, thereby smoothing the current curve and reducing the impact of noise. For example, a sliding window of 5 data points is set, and the current value is added to the current value at the previous 4 moments and then divided by 5 to obtain the smoothed current value. This can effectively suppress the instantaneous fluctuation of the current value and more accurately reflect the change trend of the load current. In order to determine the severity of the current change, it is necessary to calculate the rate of change between two adjacent current data points. The rate of change is calculated by subtracting the previous current value from the next current value and then dividing it by the time interval. For example, if the current current value is 2 amperes and the current value after 10 milliseconds is 3 amperes, the current change rate is (3-2) / 0.1=20 amperes / second. A current change rate threshold is preset, such as 10 amperes / second. When the calculated current change rate exceeds the threshold, the current is judged to be drastic; when the current change rate is lower than the threshold, the current is judged to be gentle. According to the severity of the current change, dynamically adjusting the voltage sampling frequency can improve the response speed and efficiency of the system. When the current changes drastically, it indicates that the load state changes greatly, and it is necessary to collect voltage data more frequently to capture the voltage change. At this time, the voltage sampling frequency is increased to the preset maximum value, such as 10kHz. When the current changes slowly, the load state is relatively stable, and the voltage sampling frequency can be reduced to save system resources. At this time, the voltage sampling frequency is reduced to the preset minimum value, such as 1kHz. The voltage sensor converts the analog voltage signal into a digital signal through the ADC module. The resolution and sampling rate of the ADC module affect the accuracy and speed of voltage measurement. For example, a 12-bit ADC module can divide the voltage range of 0-5 volts into 4096 levels with a resolution of 5 / 4096=22mV. Higher resolution can measure voltage values ​​more accurately. The Kalman filter algorithm is a commonly used data filtering algorithm that can effectively remove high-frequency noise in the voltage sampling data and obtain a smooth voltage curve. The core idea of ​​Kalman filtering is to use the state equation and observation equation of the system, combined with prior knowledge and the measurement data at the current moment, to make the best estimate of the state of the system. For example, the battery voltage can be used as the system state and the voltage sensor reading as the observation value. Through Kalman filtering, a more accurate and smoother battery voltage estimate can be obtained. The voltage change rate curve can be obtained by taking the first-order difference of the smoothed voltage curve. The calculation method of the voltage change rate is similar to the current change rate, that is, the difference between two adjacent voltage data points is divided by the time interval. The voltage change rate reflects the speed of voltage change. According to the voltage change rate at the current moment, dynamically adjusting the voltage sampling frequency at the next moment can further optimize the performance of the system. A voltage change rate threshold is preset, such as 1V / s.When the calculated voltage change rate is greater than the threshold, it means that the voltage changes quickly and the sampling frequency needs to be increased to capture the detailed changes in the voltage. For example, the sampling frequency is increased from 1kHz to 2kHz. When the voltage change rate is less than the threshold, it means that the voltage changes slowly and the sampling frequency can be reduced to save resources. For example, the sampling frequency is reduced from 2kHz to 1kHz. This strategy of dynamically adjusting the sampling frequency can take into account both measurement accuracy and system efficiency.

[0022] Step S104, using the Kalman filter algorithm to filter and fuse the internal resistance estimation value, dynamically track the change trend of the internal resistance, and adaptively adjust the filter parameters according to the working state and aging degree of the battery.

[0023] The working state data and aging degree data of the battery are obtained, the working state data includes voltage, current and temperature, and the aging degree data includes the number of cycles and capacity attenuation; the working state data and aging degree data are used as inputs of the Kalman filter to filter the estimated value of the battery internal resistance; according to the working state and aging degree of the battery, the process noise covariance matrix and the measurement noise covariance matrix of the Kalman filter are adaptively adjusted; the battery internal resistance is estimated by the recursive least squares method to obtain the internal resistance estimation value of the recursive least squares method; the internal resistance estimation value after Kalman filtering and the internal resistance estimation value of the recursive least squares method are weightedly fused, and the weight of the weighted fusion is dynamically adjusted according to the confidence of each estimation value to obtain the fused internal resistance estimation value; the fused internal resistance estimation value is subjected to moving average filtering to obtain the long-term change trend of the internal resistance; if the change rate or amplitude of the internal resistance exceeds the preset threshold, an alarm is triggered; according to the change trend of the internal resistance, the charge and discharge control strategy of the battery management system is adjusted; the internal resistance estimation value and the change trend are stored with the timestamp in the time series database of the battery health status.

[0024] Specifically, the working state data of the battery, such as voltage, current and temperature, and the aging degree data, such as the number of cycles and capacity attenuation, are obtained. These data can be collected by the battery management system (BMS). For example, the built-in voltage sensor in the BMS can measure the battery voltage in real time, the current sensor can measure the battery charge and discharge current, and the temperature sensor can monitor the battery temperature. The number of cycles and capacity attenuation can be obtained through the counter and capacity estimation algorithm of the BMS. These data will be used as inputs to the Kalman filter to estimate the internal resistance of the battery. The Kalman filter is an algorithm that can effectively estimate the system state from noisy measurement data. Here, it is used to estimate the internal resistance of the battery. The estimated value of the battery internal resistance is easily affected by measurement noise. Kalman filtering can help reduce the influence of noise and obtain a more accurate internal resistance estimate. For example, suppose the true value of the battery internal resistance is 0.5Ω at a certain moment, but due to the existence of measurement noise, the internal resistance measured by the BMS is 0.6Ω. The Kalman filter can combine the previous internal resistance estimate and the current measurement value to estimate the internal resistance that is closer to the true value, such as 0.52Ω. In order to improve the accuracy of internal resistance estimation, it is necessary to adaptively adjust the parameters of the Kalman filter according to the working state and aging degree of the battery. For example, when the battery is in a high-rate charge and discharge state, the current changes dramatically and the measurement noise may also be large. At this time, the measurement noise covariance matrix of the Kalman filter can be increased. When the battery ages more and the capacity decays more, the change of internal resistance may be more dramatic. At this time, the process noise covariance matrix can be increased. Assume that the measurement noise covariance matrix of a new battery is 0.01, and after the number of cycles reaches 500, the value is adjusted to 0.02 to adapt to the increase in noise caused by battery aging. In addition to Kalman filtering, other algorithms can be used to estimate the internal resistance of the battery, such as the recursive least squares method. In order to obtain a more accurate and stable internal resistance estimation result, the estimated values ​​of different algorithms can be weighted and fused. The distribution of weights can be dynamically adjusted according to the confidence of each estimated value. For example, when the estimated value of the Kalman filter fluctuates less, it can be given a higher weight; otherwise, its weight can be reduced. Assume that the internal resistance estimated by the Kalman filter is 0.5Ω with a confidence level of 9, while the internal resistance estimated by the recursive least squares method is 0.6Ω with a confidence level of 8. Then the fused internal resistance value is 0.5×9+0.6×8=0.504Ω. In order to remove short-term disturbances in the internal resistance estimate, the fused internal resistance estimate can be subjected to moving average filtering. For example, a sliding window of 5 data points can be used for moving average filtering. This means that the internal resistance estimate at the current moment will be averaged with the estimate at the previous 4 moments to obtain a smoothed internal resistance value. This helps to extract the long-term change trend of the internal resistance. For example, if the internal resistance estimates at the past 5 moments are 0.5Ω, 0.52Ω, 0.51Ω, 0.53Ω, and 0.55Ω, respectively, the internal resistance value after moving average filtering is 0.522Ω.In order to detect battery abnormalities in time, it is necessary to monitor the changes in internal resistance. The normal range threshold of internal resistance change can be preset according to the battery model and historical data. If the change rate or amplitude of the internal resistance value exceeds the threshold, an alarm is triggered, indicating that the battery needs to be maintained or replaced. For example, if the normal change rate threshold of the internal resistance of a certain model of battery is set to 0.01Ω / day, and the current measured internal resistance change rate is 0.02Ω / day, an alarm is triggered. The battery management system can dynamically adjust the charge and discharge control strategy according to the change trend of the internal resistance value. For example, if the internal resistance value continues to increase, indicating that the battery aging is aggravated, the charge and discharge rate can be reduced or the SOC usage range can be narrowed to slow down the increase rate of internal resistance and extend the cycle life of the battery. Assuming that the internal resistance of a battery has continued to increase in the past month, the maximum charge and discharge current can be reduced from 5A to 4A. Finally, the results of internal resistance estimation and trend analysis are stored in the battery health status time series database together with the corresponding timestamp. These data can provide data support for battery health status assessment and remaining life prediction. For example, these data can be used to establish a relationship model between battery internal resistance and parameters such as cycle number and capacity decay to predict the remaining life of the battery.

[0025] Step S105, combining the operating parameters including the battery temperature, current, and SOC, to establish a multivariate regression model of the internal resistance and the operating parameters, and training and optimizing the multivariate regression model through a machine learning method to achieve online adaptive correction of the internal resistance.

[0026] The real-time temperature, current and SOC operating parameters of the battery are obtained, and the operating parameters are used as input features; a multivariate linear regression model between the battery internal resistance and the operating parameters is established; the coefficients of the multivariate linear regression model are trained using a gradient descent algorithm to minimize the mean square error between the estimated internal resistance value and the actual measured value, and an optimized multivariate linear regression model is obtained; in the training process of the multivariate linear regression model, the k-fold cross-validation method is used to evaluate the generalization performance of the model with mean square error and determination coefficient as indicators, and the overfitting problem is avoided by adjusting the complexity of the model; the trained and optimized multivariate linear regression model is deployed to the battery management system, the operating parameters are obtained in real time, and the estimated internal resistance value is calculated by substituting them into the model to realize online real-time estimation of the internal resistance; the operating parameter data is continuously collected, and when the data volume reaches the preset threshold, the multivariate linear regression model is retrained using the newly collected data, and the coefficients of the model are incrementally updated to realize the adaptive correction of the internal resistance estimation model; according to the output result of the adaptively corrected internal resistance estimation model, it is judged whether the battery internal resistance is greater than the preset threshold, and if so, it is determined that the battery needs maintenance or replacement.

[0027] Specifically, first, the real-time operating parameters of the battery need to be obtained. These parameters include temperature, current, and SOC (state of charge). For example, at a certain moment, the battery temperature is 25°C, the current is 2A, and the SOC is 80%. These data can be collected in real time by sensors in the battery management system (BMS). These parameters are selected as input features because they are closely related to the internal resistance of the battery. An increase in temperature usually leads to a decrease in internal resistance, while a large current discharge increases the internal resistance. SOC also affects the internal resistance. For example, when the SOC is low, the internal resistance tends to be higher. Next, a multivariate linear regression model between the battery internal resistance and the operating parameters needs to be established. This model can be expressed as: internal resistance = a * temperature + b * current + c * SOC + d, where a, b, c, d are the coefficients of the model. For example, suppose that after training, a = -0.01, b = 0.05, c = 0.1, d = 0.2 are obtained. Then, when the temperature is 25℃, the current is 2A, and the SOC is 80%, the estimated value of the internal resistance is -0.01 * 25 +0.05 * 2 + 0.1 * 80 + 0.2 = 8.05Ω. In order to obtain a more accurate model, the coefficients of the model need to be trained using the gradient descent algorithm. The goal of the gradient descent algorithm is to minimize the mean square error between the estimated internal resistance and the actual measured value. The actual measured value can be obtained through a dedicated internal resistance measurement device. For example, 100 sets of historical data are used for training. Through continuous iteration, the gradient descent algorithm will gradually adjust the coefficients of the model to make the mean square error smaller and smaller. In order to evaluate the generalization performance of the model and prevent overfitting, the k-fold cross validation method is needed. For example, the data set is divided into 5 parts, 4 of which are used to train the model each time, and 1 is used for verification. The final model performance indicator is the average of the 5 verification results. Commonly used indicators include mean square error and determination coefficient. If the model is too complex, it is easy to overfit, resulting in good performance on the training set but poor performance on the test set. Overfitting can be avoided by adjusting the complexity of the model, such as reducing the number of features or adding regularization terms. The trained model will be deployed to the battery management system for real-time estimation of internal resistance. The BMS will collect operating parameters such as temperature, current, and SOC in real time, and substitute them into the model to calculate the estimated internal resistance. For example, when the BMS collects a temperature of 30°C, a current of 1A, and an SOC of 70%, it will immediately calculate the current estimated internal resistance. In order to maintain the accuracy of the model, it is necessary to continuously collect operating parameter data and regularly retrain the model with new data. When the amount of data reaches a preset threshold, such as 1,000 sets of new data, the coefficients of the model will be incrementally updated using these new data. This allows the model to adapt to battery aging and operating condition changes and improves estimation accuracy. Finally, the battery management system will determine the health status of the battery based on the adaptively corrected estimated internal resistance.If the estimated internal resistance is greater than a preset threshold, such as 1Ω, the battery is considered to need maintenance or replacement. For example, if the internal resistance of a battery continues to rise and exceeds the threshold, the BMS will issue an alarm to prompt the user to perform maintenance or replacement in time. This can effectively avoid battery failure and extend the battery life.

[0028] Step S106, statistically analyzing the calculated internal resistance value, extracting characteristic parameters including the internal resistance mean, variance, and change rate, combining the empirical rules in the expert knowledge base, comprehensively judging the health status of the battery, and generating a health assessment report and maintenance recommendations.

[0029] The voltage and current data of the battery during the charge and discharge process are obtained, and the internal resistance of the battery is calculated according to Ohm's law; the battery internal resistance data is statistically analyzed and processed to obtain the characteristic parameters of the mean internal resistance, the variance of the internal resistance and the rate of change of the internal resistance, wherein the mean internal resistance reflects the overall level of the battery internal resistance, the variance of the internal resistance reflects the degree of dispersion of the battery internal resistance, and the rate of change of the internal resistance reflects the trend of the battery internal resistance over time; the characteristic parameters of the mean internal resistance, the variance of the internal resistance and the rate of change of the internal resistance are input into the pre-built battery health status assessment model, which is based on historical data and expert experience and is trained using the support vector machine algorithm; if the mean internal resistance and the variance of the internal resistance exceed If the internal resistance value exceeds the normal range set in the battery health status assessment model, and the internal resistance change rate is greater than the threshold set in the battery health status assessment model, the battery health status is judged to be sub-healthy; otherwise, the battery health status is judged to be healthy; according to the judged battery health status, a battery health assessment report is automatically generated. The battery health assessment report adopts PDF format, including basic battery information, internal resistance characteristic parameters, health status assessment results and subsequent usage recommendations; the battery internal resistance value, characteristic parameters, battery health status and battery health assessment report information are stored in the MySQL database, and the data addition, deletion, modification and query operations are implemented through SQL statements.

[0030] Specifically, first, it is necessary to obtain the battery charge and discharge data. These data include voltage and current, which can be collected in real time through sensors in the battery management system (BMS). For example, during a certain discharge process, the voltage and current data are collected every second for ten minutes. These data will be used to calculate the internal resistance of the battery. According to Ohm's law, the internal resistance of the battery can be calculated by voltage drop and current. For example, at a certain moment, the battery voltage is 4.1V, the current is 1A, and the open circuit voltage of the battery is 4.2V, then the internal resistance of the battery is (4.2V-4.1V) / 1A=0.1Ω. The calculated internal resistance value is statistically analyzed to obtain characteristic parameters such as the internal resistance mean, internal resistance variance, and internal resistance change rate. For example, 600 internal resistance values ​​collected within ten minutes are statistically analyzed. Assume that the calculated internal resistance mean is 0.12Ω, which represents the overall level of the battery internal resistance. The internal resistance variance is 0.0001Ω², which reflects the degree of dispersion of the internal resistance. The smaller the variance, the more stable the internal resistance value. The internal resistance change rate is 0.001Ω / minute, which indicates the change trend of the internal resistance over time. A large change rate may mean that the battery is abnormal. These characteristic parameters will be input into the pre-built battery health assessment model. The model is trained using the support vector machine algorithm based on historical data and expert experience. For example, a large number of battery internal resistance characteristic parameters and corresponding health status data can be collected and used to train the support vector machine model. The battery health assessment model will determine the health status of the battery based on the input characteristic parameters. For example, suppose the normal internal resistance mean range set by the model is 0.08Ω to 0.15Ω, the normal internal resistance variance range is 0.00005Ω² to 0.0002Ω², and the internal resistance change rate threshold is 0.002Ω / minute. If the internal resistance mean of a battery is 0.18Ω, which is beyond the normal range, and the internal resistance variance is 0.0003Ω², which is also beyond the normal range, and the internal resistance change rate is 0.003Ω / minute, which is greater than the threshold, then the model will determine that the battery is in a sub-healthy state. If the internal resistance mean and internal resistance variance are both within the normal range, and the internal resistance change rate is less than the threshold, the battery is judged to be in a healthy state. Based on the evaluation results, the system automatically generates a battery health evaluation report. The report is in PDF format and includes basic battery information (e.g., battery model, production date, etc.), internal resistance characteristic parameters, health status evaluation results, and subsequent usage recommendations. For example, for batteries in a sub-healthy state, the report may recommend that users reduce the frequency of battery use or perform maintenance. All relevant information, including battery internal resistance values, characteristic parameters, battery health status, and battery health evaluation reports, are stored in the MySQL database. SQL statements can be used to add, delete, modify, and query data. For example, SQL statements can be used to query the health status of all batteries within a specific time period, or to query the historical internal resistance data of a battery. This facilitates data management and analysis and can provide a basis for battery maintenance and replacement.Through this battery health status assessment method based on internal resistance characteristic parameters and support vector machine, the battery health status can be monitored in real time, potential problems can be discovered in time, and corresponding maintenance suggestions can be provided, thereby effectively extending the battery life and improving the battery safety.

[0031] The above are only preferred embodiments of one or more embodiments of this specification and are not intended to limit one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this specification should be included in the scope of protection of one or more embodiments of this specification.

Claims

1. A battery health status assessment method based on BMS, characterized in that: The method comprises: Obtain multiple sets of voltage sampling data of the battery under different load currents, calculate the corresponding internal resistance estimation value based on the voltage sampling data, build a mathematical model between internal resistance, current and voltage, and determine the model parameters through curve fitting method; preset different voltage sampling ranges and resolutions for different load current intervals, and when the load current changes, determine the current interval to which it belongs, and dynamically switch to the corresponding range and resolution; when the current changes drastically, increase the voltage sampling frequency to obtain more data points, and when the current changes slowly, reduce the sampling frequency to reduce data redundancy; use the Kalman filter algorithm to filter and fuse the internal resistance estimation value, dynamically track the change trend of the internal resistance, and adaptively adjust the filter parameters according to the working state and aging degree of the battery; combine the operating parameters including battery temperature, current, and SOC to establish a multivariate regression model of internal resistance and operating parameters, train and optimize the multivariate regression model through machine learning methods, and realize online adaptive correction of internal resistance; perform statistical analysis on the calculated internal resistance value, extract characteristic parameters including internal resistance mean, variance, and change rate, and combine the empirical rules in the expert knowledge base to comprehensively judge the health status of the battery, and generate a health assessment report and maintenance recommendations.

2. The method according to claim 1, characterized in that The method comprises: obtaining multiple groups of voltage sampling data of the battery under different load currents, calculating the corresponding internal resistance estimation value according to the voltage sampling data, constructing a mathematical model between the internal resistance, current and voltage, and determining the model parameters by a curve fitting method, including: A1. Obtain the load current variation range of the battery, and design at least two different current values ​​according to the load current variation range to perform a discharge test; A2. For each current value, collect the voltage data of the battery at the current value to obtain multiple sets of current-voltage data pairs; A3. According to Ohm's law, using each set of current-voltage data pairs, calculate the estimated internal resistance value of the battery under the working condition to obtain multiple estimated internal resistance values; A4. Establishing a polynomial function relationship model among internal resistance, current and voltage according to multiple internal resistance estimation values, multiple current values ​​and corresponding voltage data; A5. Use the least square method to identify the parameters of the polynomial function relationship model and obtain the fitted internal resistance estimation model; A6. Obtain a preset data set, where the data set includes multiple sets of current-voltage data pairs; A7. Use the leave-one-out cross-validation method to divide the data set into a training set and a test set; A8. Use the training set to train the internal resistance estimation model, and use the test set to test the internal resistance estimation model to obtain the root mean square error and determination coefficient; If the root mean square error is less than a first preset threshold value and the determination coefficient is greater than a second preset threshold value, embedding the internal resistance estimation model into the battery management system; If the RMS error is greater than or equal to the first preset threshold, or the coefficient of determination is less than or equal to the second preset threshold, return to step A5; A9. During operation, the battery current and voltage data are collected in real time, substituted into the internal resistance estimation model, and the battery internal resistance under the current working conditions is calculated; A10. Dynamically adjust the battery charging and discharging strategy according to changes in the battery's internal resistance.

3. The method according to claim 1, characterized in that The method presets different voltage sampling ranges and resolutions for different load current intervals, and when the load current changes, determines the current interval to which it belongs, and dynamically switches to the corresponding range and resolution, including: Obtain the real-time value of the load current sensor, compare it with multiple preset current interval thresholds, and determine the interval range to which the load current belongs; If the load current is within a preset current interval, the voltage sampling range and ADC resolution parameters corresponding to the interval are searched from the pre-established current interval-sampling parameter mapping table; The voltage sampling range and ADC resolution parameters found are sent to the voltage sampling module through the I2C bus interface to dynamically adjust the amplification factor of its sampling circuit and the number of ADC conversion bits; During the voltage sampling process, the value change of the load current sensor is continuously monitored at a preset time interval to determine whether the load current crosses the boundary of the current interval; If the load current crosses the boundary of the current interval, the load current interval is re-determined and the parameter configuration of the voltage sampling module is updated; Collect voltage sampling data according to a preset time window, and divide the sampling data into multiple continuous sub-intervals according to the changes of the current interval and sampling parameters in each time window; The voltage sampling data in each sub-interval is subjected to denoising and filtering, and the sampling values ​​are linearly scaled according to the voltage sampling range and ADC resolution parameters of the sub-interval to obtain a unified physical voltage value; According to the average current value of each sub-interval and the corrected voltage sampling sequence, the average power and cumulative energy consumption in the sub-interval are calculated; The average power and cumulative energy consumption of each sub-interval are accumulated to obtain the total power and total energy consumption of the load in the entire time window; Compare the total power with the preset overload threshold, and compare the total energy consumption with the preset over-limit threshold to determine whether the load operation status is abnormal; If power overload or energy consumption exceeds the limit is detected, an alarm signal is triggered and the preset overload protection control logic is started.

4. The method according to claim 1, characterized in that When the current changes dramatically, the voltage sampling frequency is increased to obtain more data points; when the current changes slowly, the sampling frequency is reduced to reduce data redundancy, including: Acquire current data from the current sensor, perform sliding average filtering on the current data, and obtain a smooth current curve; Calculate the change rate between two adjacent current data points. If the current change rate exceeds a preset threshold, it is judged that the current changes violently; if the current change rate is lower than the preset threshold, it is judged that the current changes gently. If it is determined that the current changes dramatically, the voltage sampling frequency is increased to a preset maximum value; If it is judged that the current changes slowly, the voltage sampling frequency is reduced to a preset minimum value; Obtain voltage sampling data from the voltage sensor through the ADC module; The Kalman filter algorithm is used to filter the voltage sampling data to obtain a smooth voltage curve. The Kalman filter estimates the optimal value of the voltage through recursion and removes high-frequency noise. Perform the first-order difference on the smoothed voltage curve to obtain the voltage change rate curve; Calculate the voltage change rate at the current moment, compare it with the preset voltage change rate threshold, and dynamically adjust the voltage sampling frequency at the next moment; If the voltage change rate is greater than the preset threshold, the sampling frequency is increased; If the voltage change rate is less than a preset threshold, the sampling frequency is reduced.

5. The method according to claim 1, characterized in that The Kalman filter algorithm is used to filter and fuse the internal resistance estimation value, dynamically track the change trend of the internal resistance, and adaptively adjust the filter parameters according to the working state and aging degree of the battery, including: Obtaining the working status data and aging degree data of the battery, the working status data includes voltage, current and temperature, and the aging degree data includes the number of cycles and capacity attenuation; The working state data and the aging degree data are used as the input of the Kalman filter to filter the estimated value of the battery internal resistance; Adaptively adjust the process noise covariance matrix and measurement noise covariance matrix of the Kalman filter according to the working state and aging degree of the battery; The internal resistance of the battery is estimated by using the recursive least square method to obtain the internal resistance estimation value of the recursive least square method; The internal resistance estimation value after Kalman filtering is weightedly fused with the internal resistance estimation value obtained by the recursive least square method, and the weight of the weighted fusion is dynamically adjusted according to the confidence of each estimation value to obtain the fused internal resistance estimation value; Perform moving average filtering on the fused internal resistance estimation value to obtain the long-term change trend of the internal resistance; If the rate or amplitude of change of the internal resistance exceeds the preset threshold, an alarm is triggered; Adjust the charge and discharge control strategy of the battery management system according to the change trend of the internal resistance; The internal resistance estimation value and change trend are stored with timestamp in a time series database of battery health status.

6. The method according to claim 1, characterized in that The method combines the operating parameters including the battery temperature, current, and SOC to establish a multiple regression model of the internal resistance and the operating parameters, trains and optimizes the multiple regression model by a machine learning method, and realizes online adaptive correction of the internal resistance, including: Obtain the real-time temperature, current and SOC operating parameters of the battery and use the operating parameters as input features; Establish a multiple linear regression model between battery internal resistance and operating parameters; The coefficients of the multivariate linear regression model are trained using the gradient descent algorithm to minimize the mean square error between the estimated internal resistance and the actual measured value, and the optimized multivariate linear regression model is obtained. In the training process of the multivariate linear regression model, the k-fold cross-validation method is used to evaluate the generalization performance of the model with mean square error and determination coefficient as indicators, and the overfitting problem is avoided by adjusting the complexity of the model; Deploy the trained and optimized multivariate linear regression model to the battery management system, obtain the operating parameters in real time, substitute them into the model to calculate the estimated internal resistance, and realize online real-time estimation of the internal resistance; Continuously collect operating parameter data. When the data volume reaches the preset threshold, use the newly collected data to retrain the multivariate linear regression model, incrementally update the model coefficients, and realize adaptive correction of the internal resistance estimation model. According to the output result of the adaptively corrected internal resistance estimation model, it is determined whether the battery internal resistance is greater than a preset threshold. If so, it is determined that the battery needs maintenance or replacement.

7. The method according to claim 1, characterized in that The calculated internal resistance value is statistically analyzed to extract characteristic parameters including the internal resistance mean, variance, and change rate, and the health status of the battery is comprehensively judged in combination with the empirical rules in the expert knowledge base to generate a health assessment report and maintenance suggestions, including: Obtain the voltage and current data during the battery charging and discharging process, and calculate the internal resistance of the battery according to Ohm's law; The battery internal resistance data is statistically analyzed to obtain the characteristic parameters of the internal resistance mean, internal resistance variance and internal resistance change rate. The internal resistance mean reflects the overall level of the battery internal resistance, the internal resistance variance reflects the discrete degree of the battery internal resistance, and the internal resistance change rate reflects the change trend of the battery internal resistance over time. Input the characteristic parameters of internal resistance mean, internal resistance variance and internal resistance change rate into a pre-built battery health status assessment model. The battery health status assessment model is trained using a support vector machine algorithm based on historical data and expert experience. If the internal resistance mean and internal resistance variance exceed the normal range set in the battery health status assessment model, and the internal resistance change rate is greater than the threshold set in the battery health status assessment model, the battery health status is judged to be a sub-healthy state; Otherwise, the battery health state is judged to be healthy; Automatically generate a battery health assessment report based on the determined battery health status. The battery health assessment report is in PDF format and includes basic battery information, internal resistance characteristic parameters, health status assessment results, and subsequent usage recommendations. The battery internal resistance, characteristic parameters, battery health status and battery health assessment report information are stored in the MySQL database, and data addition, deletion, modification and query operations are implemented through SQL statements.

8. A battery health status assessment system based on BMS, characterized in that: The system includes: a data acquisition and modeling module, which is used to obtain voltage sampling data and build a mathematical model of internal resistance, current and voltage; a dynamic sampling control module, which is used to dynamically adjust the voltage sampling range, resolution and frequency according to the change of load current; a data filtering and fusion module, which is used to filter and dynamically track the internal resistance estimation value using the Kalman filtering algorithm; an adaptive correction module, which is used to establish a multivariate regression model of internal resistance and operating condition parameters and realize online adaptive correction; a health status assessment module, which is used to analyze the internal resistance characteristic parameters and comprehensively judge the battery health status in combination with expert knowledge.

Images