Power battery soc and soh prediction and estimation method based on temperature correction

Abstract

The application discloses a kind of power battery SOC and SOH prediction and estimation method based on temperature correction, belong to new energy automobile battery management technical field. Including the following steps: S1, in low temperature condition, through sensor synchronous acquisition power battery multi-source original data, record real-time capacity, then sequentially execute cleaning, alignment, environmental feature identification and baseline calibration to original data, output standardization dataset;S2, based on standardization dataset, temperature correction is carried out, and prediction model is constructed by fusing multiple models, and the SOC and SOH prediction result is output;S3, based on prediction result, calculate three-dimensional health index, divide health grade and dynamically adjust battery operating parameter. The power battery SOC and SOH prediction and estimation method based on temperature correction described above can effectively eliminate low temperature interference, improve the estimation accuracy of SOC and SOH, and dynamically optimize the battery operating parameters, prolong the cycle life of the battery.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle battery management technology, and in particular to a method for predicting and estimating the SOC and SOH of a power battery based on temperature correction. Background Technology

[0002] In winter, under low-temperature conditions ranging from -30℃ to 5℃, the electrochemical reaction rate of power batteries decreases significantly, and the internal resistance increases, leading to a reduction in capacity utilization and directly affecting the accurate estimation of SOC (State of Charge) and SOH (State of Health). Existing technologies have significant shortcomings. For example, the existing traditional ampere-hour integration method and equivalent circuit model do not incorporate dynamic temperature compensation, relying solely on fixed parameters at room temperature for calculation, resulting in SOC estimation errors as high as 8–10% in low-temperature environments. Existing SOH estimations mostly rely on static capacity decay models, considering only single factors such as the number of cycles, and cannot adapt to the fluctuations in internal resistance and capacity decay caused by dynamic temperature changes. None of the above technologies solve the problem of accumulated SOC and SOH estimation deviations under low-temperature conditions. Long-term use can lead to inaccurate decisions by the BMS (Battery Management System), resulting in risks such as insufficient charging, falsely advertised range, or excessive energy loss. Therefore, a precise prediction and estimation scheme incorporating a temperature correction mechanism is urgently needed. Summary of the Invention

[0003] The purpose of this invention is to provide a method for predicting and estimating the SOC and SOH of a power battery based on temperature correction, thereby solving the aforementioned technical problems.

[0004] To achieve the above objectives, this invention provides a method for predicting and estimating the SOC and SOH of a power battery based on temperature correction, comprising the following steps:

[0005] S1. Data Acquisition and Preprocessing: Under low-temperature conditions, multi-source raw data of the power battery are collected synchronously through sensors, and the real-time capacity is recorded. Then, the raw data is cleaned, aligned, environmental feature identified and baseline calibrated in sequence, and a standardized dataset is output.

[0006] S2. Temperature Correction and Prediction: S21. Based on a standardized dataset, the relationship between temperature and capacity / internal resistance is established using the Arrhenius model to correct biases in SOC calculation and SOH assessment at low temperatures. Specifically, this includes:

[0007] S211, Calculate Temperature The capacity temperature correction factor Its formula is:

[0008] ;

[0009] S212. Correct the SOC calculation result, its formula for:

[0010] ;

[0011] in, This is the current corrected SOC value. The SOC value at the previous moment. For temperature Lower charge / discharge efficiency correction factor For temperature The charging and discharging current under these conditions;

[0012] S213. The weighted method is used to correct the SOH assessment results. The formula is as follows: for:

[0013] ;

[0014] in, This is the current corrected SOH value. This is the capacity weighting coefficient. This is the internal resistance weighting coefficient. For temperature The actual usable capacity of the battery;

[0015] S214. During the constant current charging phase of the battery, the capacity increment is calculated, and the peak voltage and peak height of the capacity increment curve are extracted as auxiliary characteristics of the battery's health status. The formula for calculating the capacity increment is as follows: for:

[0016] ;

[0017] in, For the increase in charging capacity, For voltage increment, For constant current charging current, For charging time increments, for The initial sampled voltage value, for The sampled voltage value at the end;

[0018] S22. Based on the different characteristics of short-term dynamic changes in SOC and long-term decay in SOH, a prediction model is constructed by integrating multiple models, specifically including:

[0019] S221. Use an LSTM model to predict short-term dynamic changes in SOC. Input the corrected SOC, voltage, current, and temperature time-series data into the LSTM model, and use the Adam optimizer with mean squared error. As a loss function, the current SOC prediction value is output, where... The calculation formula is:

[0020] ;

[0021] in, The number of training samples, This represents the actual SOC value. This is the predicted SOC value;

[0022] S222. The long-term decay of SOH is predicted by fusing the grey GM(1,1) model and the LSTM model with the entropy weight method. The grey GM(1,1) model collects data at equal time intervals. Using corrected SOH data, construct the original sequence. ,right Generate a new sequence by performing an accumulation. A first-order linear differential equation model was established, and the development coefficient was calculated using the least squares method. and gray action amount The formula for the first-order linear differential equation is:

[0023] ;

[0024] The long-term prediction formula for SOH is:

[0025] ;

[0026] in, This serves as the starting data point identifier for the original SOH sequence. To predict the number of steps; a temperature acceleration factor is introduced using the Arrhenius model. Dynamically corrected SOH attenuation prediction coefficient Among them, temperature acceleration factor and attenuation prediction coefficient The formulas are as follows:

[0027] ;

[0028] ;

[0029] in, This is for the activation energy of battery aging. The gas constant is... For reference absolute temperature, The average absolute operating temperature of the battery. The baseline decay coefficient is used; within the same prediction period, the SOH sequence output by the LSTM model is obtained. Standard deviation The grey GM(1,1) model outputs the SOH sequence. Standard deviation The information entropy of the LSTM model and the grey GM(1,1) model is calculated using the following formula:

[0030] ;

[0031] in, The information entropy of the LSTM model, The information entropy of the GM(1,1) model; for , The normalization yields the fusion weights, and the formula is as follows:

[0032] ;

[0033] The formula for the weighted fusion of SOH prediction values ​​is:

[0034] ;

[0035] S23. Output the predicted SOC and SOH results after temperature correction;

[0036] S3. Health Assessment: Based on the prediction results of S23, calculate the three-dimensional health indicators, including the SOC estimation error. SOH decay rate Temperature sensitivity Among them, temperature sensitivity The calculation formula is:

[0037] ;

[0038] in, The low temperature point room temperature point for The measured internal resistance is below. for The measured internal resistance is as follows;

[0039] The health levels are then further classified and the battery operating parameters are dynamically adjusted.

[0040] Preferably, S1 specifically includes:

[0041] S11. Voltage, current, and temperature data of the power battery are collected synchronously using voltage, current, and temperature sensors. The original sampling frequency of the sensors is set to... Simultaneously, it records the real-time battery capacity;

[0042] S12. Data cleaning uses a low-pass filter to remove high-frequency noise interference from voltage and current signals, and continuously missing durations. Long-term data segments of seconds are marked as invalid and are not included in the modeling calculations. Missing short-term data are filled using linear interpolation. The formula for linear interpolation is:

[0043] ;

[0044] in, To fill in the value, The data before the missing points is valid data. Valid data after the missing points. For the missing time point, The effective data time before the missing point. The effective data time after the missing points; outliers are removed using a statistical threshold method, and the arithmetic mean of the data within a 5-second time window is calculated. and standard deviation Remove excess Abnormal data, including the arithmetic mean and standard deviation The calculation formula is:

[0045] ;

[0046] ;

[0047] in, The number of data points within a 5-second window. For the first in the window One data point;

[0048] S13. Data alignment unifies the time step through downsampling and aggregation. Output aligned voltage, current, and temperature data;

[0049] S14. Environmental Feature Identification and Baseline Calibration: Under standard conditions, the battery is discharged from full charge to cutoff voltage at a rate of 0.33C, and the discharge capacity is recorded as the rated discharge capacity. The standard internal resistance of the battery was measured using a hybrid pulse power characteristic test method. Then change the temperature Repeat the test and record the different temperatures. Measured discharge capacity and measured internal resistance To establish an OCV-SOC benchmark correspondence table;

[0050] S15, Standardized dataset of output voltage, current, and temperature.

[0051] Preferably, S3 further includes:

[0052] S31. In the three-dimensional health indicators, the SOC estimation error The calculation formula is:

[0053] ;

[0054] in, It is the arithmetic mean. This is the predicted SOC value. The SOC reference value is obtained under standard conditions;

[0055] SOH decay rate The calculation formula is:

[0056] ;

[0057] in, To assess the duration of the evaluation period, For the evaluation period Initial SOH value, For the evaluation period SOH value at the end;

[0058] S32. Based on three-dimensional health indicators, the battery health status is quantified and divided into four health levels: Level I is excellent, Level II is good, Level III is caution required, and Level IV requires replacement. The specific standards are as follows:

[0059] Level I: , , ;

[0060] Level II: , , ;

[0061] Level III: , , ;

[0062] Level IV: , , ;

[0063] S33. Feedback the health level to the power battery management system. Based on the health level, adjust the operating parameters, including the maximum allowable charging current, the maximum allowable discharging current, the SOC range, and the temperature alarm threshold. The adjustment criteria for the control strategy mapping table corresponding to the health level are as follows:

[0064] Level I: Maximum allowable charging current 1C, maximum allowable discharging current 2C, SOC range 20%~100%, temperature alarm threshold -10℃;

[0065] Level II: Maximum allowable charging current 0.8C, maximum allowable discharging current 1.5C, SOC range 20%~95%, temperature alarm threshold -5℃;

[0066] Level III: Maximum allowable charging current 0.5C, maximum allowable discharging current 1C, SOC range 30%~90%, temperature alarm threshold 0℃;

[0067] Level IV: Maximum allowable charging current 0.2C, maximum allowable discharging current 0.5C, SOC range 50%~80%, temperature alarm threshold 10℃.

[0068] Preferably, the first-order low-pass filter in S12 is a Butterworth low-pass filter with a fixed cutoff frequency of 10Hz.

[0069] Preferably, the time step in S13 The value range is 0.1s to 10s. Number of sampling points within The calculation formula is:

[0070] .

[0071] Preferably, during the establishment of the OCV-SOC benchmark correspondence table in S14, the battery is charged and discharged at a constant current of 0.05C. After each 10% decrease in SOC, the battery is left to stand for 1 hour, and the corresponding OCV value is recorded until an OCV-SOC benchmark correspondence table covering the 0%~100% SOC range is established.

[0072] Preferably, when calculating the capacity increment in S214, the fluctuation range of the battery charging current during the constant current charging stage is considered. To prevent current fluctuations from affecting Interference in curve feature extraction.

[0073] Preferably, the LSTM model in S221 includes an input layer, two hidden layers, and an output layer. The input layer inputs four time-series data features: SOC, voltage, current, and temperature, after correction at the previous time step. Each hidden layer has 64 to 128 neurons. The ReLU activation function is used between the input layer and the hidden layer, and between the hidden layer and the output layer. The output layer outputs the current predicted SOC value.

[0074] Preferably, S1 outputs a standardized dataset of voltage, current, and temperature through the data acquisition and preprocessing module, providing noise-free and unbiased input data for S2; S2 outputs the real-time predicted value of SOC and the long-term decay trend of SOH after temperature correction through the temperature correction and prediction module; S3 outputs a battery health level report and operating parameter adjustment instructions through the health assessment module, realizing visualized management and active protection of battery health status.

[0075] Therefore, the above-mentioned temperature-corrected method for predicting and estimating the SOC and SOH of power batteries has the following beneficial effects:

[0076] 1. By using a temperature correction mechanism and preprocessing the data, low-temperature interference is effectively eliminated, reducing SOC estimation error and SOH assessment deviation.

[0077] 2. The SOC short-term forecast uses the LSTM model to accurately track dynamic changes, while the SOH long-term forecast uses the gray GM(1,1) model to lock in the macro trend. Then, the short-term fluctuation estimation capability of LSTM is integrated through the entropy weight method, which takes into account both long-term robustness and short-term sensitivity, making it more reliable than a single model.

[0078] 3. Dynamically adjust operating parameters according to health level to avoid excessive battery wear under harsh operating conditions, extend cycle life, and mitigate safety risks through temperature alarms.

[0079] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0080] Figure 1 This is an overall flowchart of the temperature-corrected power battery SOC and SOH prediction and estimation method of the present invention;

[0081] Figure 2 This is a comparison diagram of the capacity increment curve characteristics of the present invention;

[0082] Figure 3 This is a module relationship diagram of the present invention. Detailed Implementation

[0083] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0084] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0085] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0086] like Figures 1-3 As shown, this invention provides a method for predicting and estimating the State of Charge (SOC) and State of Health (SOH) of a power battery based on temperature correction. Using a lithium-ion power battery as the test object, and targeting a low-temperature operating condition of -30℃ to 5℃, the method achieves accurate prediction and health management of SOC and SOH through the logical execution of data acquisition and preprocessing, temperature correction and prediction, and health assessment. Figure 1 As shown, the overall technical solution includes the following steps: a temperature-corrected method for predicting and estimating the SOC and SOH of a power battery, including the following steps: S1, Data acquisition and preprocessing: Under low-temperature conditions, multi-source raw data of the power battery are collected synchronously through sensors, real-time capacity is recorded, and then the raw data is sequentially cleaned, aligned, environmental feature identified, and baseline calibrated to output a standardized dataset; S2, Temperature correction and prediction: Temperature correction is performed based on the standardized dataset, multiple models are fused to construct a prediction model, and the SOC and SOH prediction results are output; S3, Health assessment: Three-dimensional health indicators are calculated based on the prediction results, health levels are classified, and battery operating parameters are dynamically adjusted. All links work together to manage the battery status under low temperature conditions.

[0087] The purpose of S1 is to obtain the original parameters of the battery under low-temperature conditions, providing a reliable data source for subsequent processing and modeling. Specifically, this includes: S11, using voltage sensors, current sensors, and temperature sensors to simultaneously collect the voltage, current, and temperature data of the power battery, with the original sampling frequency of the sensors set to... ,like The frequency can be set to 100Hz, and the real-time battery capacity is recorded simultaneously; S12, data cleaning uses a low-pass filter to remove high-frequency noise interference in the voltage and current signals, and records continuous missing durations. Long-term data segments of seconds are marked as invalid and are not included in the modeling calculations. Missing short-term data are filled using linear interpolation. The formula for linear interpolation is:

[0088] ;

[0089] in, To fill in the value, The data before the missing points is valid data. Valid data after the missing points. For the missing time point, The effective data time before the missing point. The effective data time after the missing points; outliers are removed using a statistical threshold method, and the arithmetic mean of the data within a 5-second time window is calculated. and standard deviation Remove excess Abnormal data, including the arithmetic mean and standard deviation The calculation formula is:

[0090] ;

[0091] ;

[0092] in, This represents the number of data points within a 5-second window. For the first in the window One data point; S13, Data alignment is achieved by downsampling and aggregation to unify the time step. Output aligned voltage, current, and temperature data; S14, Environmental Feature Identification and Baseline Calibration, involves discharging the battery from full charge to cutoff voltage at a rate of 0.33C under a standard environment of 25℃±2℃, and recording the discharge capacity as the rated discharge capacity. The standard internal resistance of the battery was measured using a hybrid pulse power characteristic test method. Then change the temperature Repeat the test, temperature Set temperatures of -30℃, -20℃, -10℃, 0℃, and 5℃ respectively, and record the results. Measured discharge capacity and measured internal resistance To establish an OCV-SOC reference correspondence table, a standardized dataset of S15, output voltage, current, and temperature.

[0093] S2 achieves accurate estimation of SOC and SOH through temperature correction to eliminate low-temperature interference and multi-model fusion prediction. Specifically, it includes: S21, based on the standardized dataset output by S15, establishing the relationship between temperature, capacity, and internal resistance through the Arrhenius model, correcting the calculation bias of SOC and the evaluation bias of SOH at low temperatures. Specifically, it includes: S211, calculating temperature... The capacity temperature correction factor Its formula is:

[0094] ;

[0095] In a temperature-controlled experimental chamber, a 50Ah lithium-ion battery was tested at different temperatures. The test data for its internal resistance and capacity are shown in Table 1.

[0096] Table 1. Test data on battery internal resistance and capacity changes with temperature.

[0097]

[0098] S212. Correct the SOC calculation result, its formula for:

[0099] ;

[0100] in, This is the current corrected SOC value. The SOC value at the previous moment. For temperature Lower charge / discharge efficiency correction factor For temperature The charging and discharging current under the given conditions; S213, the weighted method is used to correct the SOH evaluation results, the formula is as follows: for:

[0101] ;

[0102] in, This is the current corrected SOH value. This is the capacity weighting coefficient. This is the internal resistance weighting coefficient. For temperature The actual usable capacity of the battery is determined; S214, during the constant current charging phase of the battery, the capacity increment is calculated and the peak voltage and peak height of the capacity increment curve are extracted as auxiliary characteristics of the battery's health status, such as... Figure 2 As shown, the horizontal axis represents the voltage increment. The vertical axis represents the capacity increment. , Figure 2 Peak height of the healthy battery curve Corresponding peak voltage Peak height of aging battery curve Corresponding peak voltage Peak value differences can help determine the degree of battery aging, including the formula for calculating capacity increment. for:

[0103] ;

[0104] in, For constant current charging current, For charging time increments, for The initial sampled voltage value, for The sampled voltage value at the end; S22, Based on the different characteristics of short-term dynamic changes in SOC and long-term decay in SOH, a prediction model is constructed by fusing multiple models, specifically including: S221, Using an LSTM model to predict short-term dynamic changes in SOC, inputting the corrected SOC, voltage, current, and temperature time-series data into the LSTM model, and using the Adam optimizer with mean squared error... As a loss function, the current SOC prediction value is output, where... The calculation formula is:

[0105] ;

[0106] in, The number of training samples, This represents the actual SOC value. S222: The long-term decay of SOH is predicted using the entropy weight method, which combines the grey GM(1,1) model and the LSTM model. The grey GM(1,1) model collects data at equal time intervals. Using corrected SOH data, construct the original sequence. ,right Generate a new sequence by performing an accumulation. A first-order linear differential equation model was established, and the development coefficient was calculated using the least squares method. and gray action amount The formula for the first-order linear differential equation is:

[0107] ;

[0108] The long-term prediction formula for SOH is:

[0109] ;

[0110] in, This serves as the starting data point identifier for the original SOH sequence. To predict the number of steps; a temperature acceleration factor is introduced using the Arrhenius model. Dynamically corrected SOH attenuation prediction coefficient Among them, temperature acceleration factor and attenuation prediction coefficient The formulas are as follows:

[0111] ;

[0112] ;

[0113] in, This is for the activation energy of battery aging. The gas constant is... For reference absolute temperature, The average absolute operating temperature of the battery. The baseline decay coefficient is used; within the same prediction period, taking week 6 as an example, the output of the LSTM model is used. The sequence is Its standard deviation ,Pick , The predicted SOH for week 6 is 93.7%, and the SOH for week 7 is 91.1%. The grey GM(1,1) model outputs... The sequence is Its standard deviation The formulas for calculating the information entropy of the LSTM model and the grey GM(1,1) model are as follows:

[0114] ;

[0115] in, The information entropy of the LSTM model, The information entropy of the GM(1,1) model; for , The normalization yields the fusion weights, and the formula is as follows:

[0116] ;

[0117] The formula for the weighted fusion of SOH prediction values ​​is:

[0118] ;

[0119] S23. Output the predicted SOC and SOH results after temperature correction.

[0120] Based on the prediction results of S3, battery health management is achieved through health level assessment and operating parameter adjustment. Specifically, this includes: S31, calculating three-dimensional health indicators based on the prediction results output by S23. The three-dimensional health indicators include SOC estimation error. SOH decay rate Temperature sensitivity Among them, SOC estimation error The calculation formula is:

[0121] ;

[0122] in, It is the arithmetic mean. This is the predicted SOC value. The SOC reference value is measured under standard conditions; SOH decay rate. The calculation formula is:

[0123] ;

[0124] in, To assess the duration of the evaluation period, For the evaluation period Initial SOH value, For the evaluation period SOH value at the end; temperature sensitivity The calculation formula is:

[0125] ;

[0126] in, This refers to low-temperature points, such as -20℃, for testing. The ambient temperature is the normal temperature point. , for The measured internal resistance is below. for The measured internal resistance is as follows; S32, the battery health status is quantified based on three-dimensional health indicators and divided into 4 health levels, where Level I is excellent, Level II is good, Level III is noteworthy, and Level IV is replacement. The specific standards are as follows: Level I: , , Level II: , , Level III: , , Level IV: , , S33. Feedback the health level to the power battery management system. Based on the health level, adjust the operating parameters, including the maximum allowable charging current, maximum allowable discharging current, SOC range, and temperature alarm threshold. The adjustment standards for the control strategy mapping table corresponding to the health level are as follows: Level I: Maximum allowable charging current 1C, maximum allowable discharging current 2C, SOC range 20%~100%, temperature alarm threshold -10℃; Level II: Maximum allowable charging current 0.8C, maximum allowable discharging current 1.5C, SOC range 20%~95%, temperature alarm threshold -5℃; Level III: Maximum allowable charging current 0.5C, maximum allowable discharging current 1C, SOC range 30%~90%, temperature alarm threshold 0℃; Level IV: Maximum allowable charging current 0.2C, maximum allowable discharging current 0.5C, SOC range 50%~80%, temperature alarm threshold 10℃.

[0127] Furthermore, the first-order low-pass filter in S12 is a Butterworth low-pass filter with a fixed cutoff frequency of 10Hz. The time step in S13... The value range is 0.1s to 10s. Number of sampling points within The calculation formula is:

[0128] .

[0129] During the establishment of the OCV-SOC benchmark correspondence table in S14, constant current charging and discharging at 0.05C is used. After each 10% decrease in battery SOC, the battery is left to rest for 1 hour, and the corresponding OCV value is recorded until an OCV-SOC benchmark correspondence table covering the 0%~100% SOC range is established. In S214, when calculating the capacity increment, the fluctuation range of the battery charging current during the constant current charging phase is considered. To prevent current fluctuations from affecting Interference in curve feature extraction. The LSTM model in S221 includes an input layer, two hidden layers, and an output layer. The input layer takes into account the corrected SOC, voltage, current, and temperature features from the previous time step. Each hidden layer has 64-128 neurons. ReLU activation functions are used between the input and hidden layers, and between the hidden and output layers. The output layer outputs the current predicted SOC value. Figure 3 As shown, S1 outputs a standardized dataset of voltage, current, and temperature through the data acquisition and preprocessing module, providing noise-free and bias-free input data for S2; S2 outputs the real-time predicted value of SOC and the long-term decay trend of SOH after temperature correction through the temperature correction and prediction module; S3 outputs a battery health level report and operating parameter adjustment instructions through the health assessment module, realizing visualized management and active protection of battery health status. To verify the effectiveness of this invention, this embodiment designs a comparative and simulation experiment. The temperature of the temperature control box is set to -10°C, and three groups of power lithium batteries of the same model and capacity of 50Ah are selected. The SOC is estimated using the traditional ampere-hour integration method and the method of this invention. The results measured by a high-precision charge and discharge test equipment under a standard environment of 25°C are used as the SOC reference true value. The specific results are shown in Table 2.

[0130] Table 2 Comparison of Mean Absolute Error (MAE) for SOC Estimation at -10°C

[0131]

[0132] The experimental data above show that, under the harsh low-temperature environment of -10°C, this invention, by introducing a temperature correction mechanism, significantly reduces the average error to 2.1%, and improves the overall estimation accuracy by 75.4%. Furthermore, in a simulated prediction of SOH degradation at 0°C for 300 cycles, after 100 cycles, the invention predicted an SOH of 95.2%, while the measured SOH was 95.5%, with an error of 0.3%; after 200 cycles, the invention predicted an SOH of 88.1%, while the measured SOH was 87.5%, with an error of 0.6%; after 300 cycles, the invention predicted an SOH of 78.5% and provided an early warning of the capacity "plummeting" trend. The measured SOH subsequently dropped rapidly to 76.1% after 320 cycles, proving that this invention can not only accurately estimate the current SOH but also effectively predict long-term degradation trends, providing a crucial basis for battery life prediction.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting and estimating the SOC and SOH of a power battery based on temperature correction, characterized in that: Includes the following steps: S1. Data Acquisition and Preprocessing: Under low-temperature conditions, multi-source raw data of the power battery are collected synchronously through sensors, and the real-time capacity is recorded. Then, the raw data is cleaned, aligned, environmental feature identified and baseline calibrated in sequence, and a standardized dataset is output. S2. Temperature Correction and Prediction: S21. Based on a standardized dataset, the relationship between temperature and capacity / internal resistance is established using the Arrhenius model to correct biases in SOC calculation and SOH assessment at low temperatures. Specifically, this includes: S211, Calculate Temperature The capacity temperature correction factor Its formula is: ； S212. Correct the SOC calculation result, its formula for: ； in, This is the current corrected SOC value. The SOC value at the previous moment. For temperature Lower charge / discharge efficiency correction factor For temperature The charging and discharging current under these conditions; S213. The weighted method is used to correct the SOH assessment results. The formula is as follows: for: ； in, This is the current corrected SOH value. This is the capacity weighting coefficient. This is the internal resistance weighting coefficient. For temperature The actual usable capacity of the battery; S214. During the constant current charging phase of the battery, the capacity increment is calculated, and the peak voltage and peak height of the capacity increment curve are extracted as auxiliary characteristics of the battery's health status. The formula for calculating the capacity increment is as follows: for: ； in, For the increase in charging capacity, For voltage increment, For constant current charging current, For charging time increments, for The initial sampled voltage value, for The sampled voltage value at the end; S22. Based on the different characteristics of short-term dynamic changes in SOC and long-term decay in SOH, a prediction model is constructed by integrating multiple models, specifically including: S221. Use an LSTM model to predict short-term dynamic changes in SOC. Input the corrected SOC, voltage, current, and temperature time-series data into the LSTM model, and use the Adam optimizer with mean squared error. As a loss function, the current SOC prediction value is output, where... The calculation formula is: ； in, The number of training samples, This represents the actual SOC value. This is the predicted SOC value; S222. The long-term decay of SOH is predicted by fusing the grey GM(1,1) model and the LSTM model with the entropy weight method. The grey GM(1,1) model collects data at equal time intervals. Using corrected SOH data, construct the original sequence. ,right Generate a new sequence by performing an accumulation. A first-order linear differential equation model was established, and the development coefficient was calculated using the least squares method. and gray action amount The formula for the first-order linear differential equation is: ； The long-term prediction formula for SOH is: ； in, This serves as the starting data point identifier for the original SOH sequence. To predict the number of steps; a temperature acceleration factor is introduced using the Arrhenius model. Dynamically corrected SOH attenuation prediction coefficient Among them, temperature acceleration factor and attenuation prediction coefficient The formulas are as follows: ； ； in, This is for the activation energy of battery aging. The gas constant is... For reference absolute temperature, The average absolute operating temperature of the battery. The baseline decay coefficient is used; within the same prediction period, the SOH sequence output by the LSTM model is obtained. Standard deviation The grey GM(1,1) model outputs the SOH sequence. Standard deviation The information entropy of the LSTM model and the grey GM(1,1) model is calculated using the following formula: ； in, The information entropy of the LSTM model. The information entropy of the GM(1,1) model; for , The normalization yields the fusion weights, and the formula is as follows: ； The formula for the weighted fusion of SOH prediction values ​​is: ； S23. Output the predicted SOC and SOH results after temperature correction; S3. Health Assessment: Based on the prediction results of S23, calculate the three-dimensional health indicators, including the SOC estimation error. SOH decay rate Temperature sensitivity Among them, temperature sensitivity The calculation formula is: ； in, The low temperature point room temperature point for The measured internal resistance is below. for The measured internal resistance is as follows; The health levels are then further classified and the battery operating parameters are dynamically adjusted.

2. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 1, characterized in that: S1 specifically includes: S11. Voltage, current, and temperature data of the power battery are collected synchronously using voltage, current, and temperature sensors. The original sampling frequency of the sensors is set to... Simultaneously, it records the real-time battery capacity. S12. Data cleaning uses a low-pass filter to remove high-frequency noise interference from voltage and current signals, and continuously missing durations. Long-term data segments of seconds are marked as invalid and are not included in the modeling calculations. Missing short-term data are filled using linear interpolation. The formula for linear interpolation is: ； in, To fill in the value, The data before the missing points is valid data. Valid data after the missing points. For the missing time point, The effective data time before the missing point. The effective data time after the missing points; outliers are removed using a statistical threshold method, and the arithmetic mean of the data within a 5-second time window is calculated. and standard deviation Remove excess Abnormal data, including the arithmetic mean and standard deviation The calculation formula is: ； ； in, The number of data points within a 5-second window. For the first in the window One data point; S13. Data alignment unifies the time step through downsampling and aggregation. Output aligned voltage, current, and temperature data; S14. Environmental Feature Identification and Baseline Calibration: Under standard conditions, the battery is discharged from full charge to cutoff voltage at a rate of 0.33C, and the discharge capacity is recorded as the rated discharge capacity. The standard internal resistance of the battery was measured using a hybrid pulse power characteristic test method. Then change the temperature Repeat the test and record the different temperatures. Measured discharge capacity and measured internal resistance To establish an OCV-SOC benchmark correspondence table; S15, Standardized dataset of output voltage, current, and temperature.

3. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 2, characterized in that: S3 also includes: S31. In the three-dimensional health indicators, the SOC estimation error The calculation formula is: ； in, It is the arithmetic mean. This is the predicted SOC value. The SOC reference value is obtained under standard conditions; SOH decay rate The calculation formula is: ； in, To assess the duration of the evaluation period, For the evaluation period Initial SOH value, For the evaluation period SOH value at the end; S32. Based on three-dimensional health indicators, the battery health status is quantified and divided into four health levels: Level I is excellent, Level II is good, Level III is caution required, and Level IV requires replacement. The specific standards are as follows: Level I: , , ; Level II: , , ; Level III: , , ; Level IV: , , ; S33. Feedback the health level to the power battery management system. Based on the health level, adjust the operating parameters, including the maximum allowable charging current, the maximum allowable discharging current, the SOC range, and the temperature alarm threshold. The adjustment criteria for the control strategy mapping table corresponding to the health level are as follows: Level I: Maximum allowable charging current 1C, maximum allowable discharging current 2C, SOC range 20%~100%, temperature alarm threshold -10℃; Level II: Maximum allowable charging current 0.8C, maximum allowable discharging current 1.5C, SOC range 20%~95%, temperature alarm threshold -5℃; Level III: Maximum allowable charging current 0.5C, maximum allowable discharging current 1C, SOC range 30%~90%, temperature alarm threshold 0℃; Level IV: Maximum allowable charging current 0.2C, maximum allowable discharging current 0.5C, SOC range 50%~80%, temperature alarm threshold 10℃.

4. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 2, characterized in that: The first-order low-pass filter in S12 is a Butterworth low-pass filter with a fixed cutoff frequency of 10Hz.

5. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 2, characterized in that: Time step in S13 The value range is 0.1s to 10s. Number of sampling points within The calculation formula is: 。 6. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 2, characterized in that: During the establishment of the OCV-SOC benchmark correspondence table in S14, the battery is charged and discharged at a constant current of 0.05C. After each 10% decrease in SOC, the battery is left to stand for 1 hour, and the corresponding OCV value is recorded until an OCV-SOC benchmark correspondence table covering the 0%~100% SOC range is established.

7. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 1, characterized in that: When calculating the capacity increment in S214, the fluctuation range of the battery charging current during the constant current charging phase is considered. To prevent current fluctuations from affecting Interference in curve feature extraction.

8. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 1, characterized in that: The LSTM model in S221 includes an input layer, two hidden layers, and an output layer. The input layer takes into account four time-series data features: the corrected SOC, voltage, current, and temperature from the previous time step. Each hidden layer has 64 to 128 neurons. The ReLU activation function is used between the input layer and the hidden layer, and between the hidden layer and the output layer. The output layer outputs the current predicted SOC value.

9. The method for predicting and estimating the SOC and SOH of a power battery based on temperature correction according to claim 1, characterized in that: S1 outputs standardized datasets of voltage, current, and temperature through the data acquisition and preprocessing module, providing noise-free and unbiased input data for S2; S2 outputs the real-time predicted value of SOC and the long-term decay trend of SOH after temperature correction through the temperature correction and prediction module; S3 outputs battery health level reports and operating parameter adjustment instructions through the health assessment module, realizing visualized management and active protection of battery health status.

Images