Battery aging classification early warning method applied to CEBMS in data loss scene

Through the combination of Kalman filter and electric heating model, the challenge of early warning of battery aging in the lost data scenario in CEBMS is solved, accurate evaluation of aging types and model parameter updates are achieved, and the stability and accuracy of the battery management system are improved. .

CN118914856APending Publication Date: 2024-11-08BEIJING INST OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410976626.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-21
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In CEBMS, how to evaluate the error level between the cloud battery model and the actual battery system in the lost data scenario, and then conduct a classification warning of battery aging is a challenge, especially because far-field wireless communication is susceptible to interference and has a high bit error rate in transmission. This results in a variety of aging types and difficulty in updating model parameters.

Method used

Using Kalman filter for data denoising, combining electrical and thermal models, by fitting the OCV-SOC curve and establishing equivalent circuit models, aging thresholds and early warning indicators are extracted, and classified aging warning and model update strategies are designed, including capacity decay , equivalent circuit model deviation and thermal model parameter warning to ensure model accuracy.

Benefits of technology

Effectively evaluate the degree of battery aging in the lost data scenario, realize the classification of aging types, guide the update of cloud BMS model parameters, improve the stability and accuracy of CEBMS, and ensure the accuracy of battery management algorithms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914856A_ABST
    Figure CN118914856A_ABST
Patent Text Reader

Abstract

The invention discloses a battery aging classification early warning method applied to a CEBMS in a data loss scene, and the method comprises the steps: S1, carrying out the denoising of original data based on a Kalman filter, and obtaining the denoised data; s2, inputting the current data into an electric model and a thermal model to obtain estimated voltage, SOC and temperature data; s3, inputting the estimated data into an AWI calculation module, inputting preset historical data into a WT calculation module to perform aging threshold extraction, constructing a classification aging early warning and model updating strategy according to the relative size of the aging threshold and the aging early warning index, and updating the aging early warning model according to the classification aging early warning and model updating strategy. According to the method, the aging degree of the battery can be evaluated and the aging types can be classified under the scene that the cloud BMS receives the lost data, so that reference and guidance are provided for parameter calibration and model updating of the cloud battery model in the CEBMS, the precision of the battery model established in the cloud BMS is ensured, and the stability and accuracy of the CEBMS are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of energy management, and in particular relates to a battery aging classification early warning method for CEBMS applied in a data loss scenario. Background Art

[0002] Under the dual challenges of the current energy crisis and environmental pollution, the urgent demand for energy conservation and new energy has driven new energy-saving vehicles represented by pure electric vehicles to become the focus of global research. In energy-saving and new energy vehicles, the on-board power battery system is regarded as the key power core, and its energy and power characteristics determine the performance of the entire electric vehicle. The working performance of the on-board power battery system is also regulated by the battery management system (BMS), which ensures the high-performance and safe operation of electric vehicles. In addition to the performance of the power battery system itself, the performance of the BMS also greatly affects the key indicators of electric vehicles such as cruising range, power output and overall safety. In addition to focusing on improving the performance of the power battery system itself, researchers also focus on the performance optimization of the BMS to improve the overall performance of electric vehicles. This comprehensive optimization work is of great significance for coping with energy challenges, reducing environmental pollution, and promoting the development of clean energy transportation.

[0003] BMS collects data such as the operating voltage, current, and surface temperature of the vehicle-mounted power battery system, and inputs these data into advanced battery management algorithms to estimate the operating status of the power battery system and control the charging and discharging of the battery system to ensure the safe and stable operation of the battery system. At present, the management algorithm based on the battery equivalent circuit model has been widely studied, but in practical applications, the computing resources of the vehicle-side BMS embedded system are limited and the operation speed is low, which directly leads to the need for the model of practical application to trade off processing speed, accuracy, and complexity. The accuracy of the practical application model is limited because it is difficult to fully present the complex battery working state under limited computing power, resulting in low control accuracy, large estimation error, and large cumulative error. In order to solve the above problems and improve the management performance of the vehicle-side BMS, it is necessary to further explore innovative algorithms and technologies to optimize the performance of the BMS and ensure the safety and performance stability of the electric vehicle battery system.

[0004] With the continuous development of battery big data platform and edge computing technology, the "cloud-edge" battery management system (CEBMS) technology based on "cloud computing" has attracted widespread attention. This technology uses far-field wireless communication technology to transmit the battery working data (such as voltage, current, temperature) and estimated battery state data (SOC) collected by the on-board BMS to the cloud BMS in real time, and performs data storage and processing and battery management in the cloud. The results of the cloud BMS state estimation are sent to the vehicle-side BMS through far-field wireless communication technology to ensure the safe and reliable operation of the vehicle battery system and achieve efficient management of the battery system. The cloud BMS can rely on the massive data and distributed large-scale computing resources of the cloud database to run advanced battery management algorithms and perform more accurate state estimation and fault diagnosis of the battery system. This architecture enables the key tasks of battery management to be carried out efficiently on the cloud BMS, thereby providing more accurate and reliable management for the vehicle battery system. The CEBMS architecture can effectively overcome the problems of limited vehicle-side computing resources and insufficient model accuracy, and improve the work efficiency and performance of electric vehicle battery system management.

[0005] In CEBMS, the cloud BMS builds a cloud battery model based on historical data in the database to estimate the battery status and implement advanced battery management algorithms. However, as the number of charge and discharge cycles of the on-board battery system increases, the battery pack will gradually age, and the cloud BMS will find it difficult to detect the occurrence of battery aging, which will directly lead to a gradual increase in the difference between the cloud battery model and the actual battery system. This deviation will directly affect the accuracy and estimation results of the cloud battery model, and ultimately affect the performance of CEBMS. In addition, the wireless communication technologies such as 5G and Wi-Fi used in the CEBMS architecture have problems such as susceptibility to interference in far-field wireless communication, high bit errors and packet loss rates during transmission. At the same time, the manifestations of aging are diverse, including capacity decline, increased internal resistance, parameter deviation, etc., and the model parameters that need to be updated for different types of aging are different. In summary, it is challenging to evaluate the error level between the cloud battery model and the actual battery system in the case of data loss in CEBMS, and then classify and warn the aging of the battery system. Therefore, it is necessary to propose a method for battery aging classification warning in the case of data loss in CEBMS. When the aging warning is triggered, the cloud BMS is guided to update the model and calibrate the parameters of the corresponding parts to ensure that the accuracy of the cloud battery model in CEBMS meets the requirements. Summary of the invention

[0006] The present invention aims to solve the technical problems existing in the background technology and provides a battery aging classification warning method for CEBMS in a data loss scenario.

[0007] In order to solve the technical problem, the technical solution of the present invention is:

[0008] A battery aging classification warning method for CEBMS in a data loss scenario, the method comprising:

[0009] S1: De-noise the original data based on the Kalman filter to obtain the de-noised voltage, current, SOC and temperature data;

[0010] S2: Input the current data into the electrical model and the thermal model to obtain the estimated voltage, SOC and temperature data;

[0011] S3: The estimated data is input into the AWI calculation module, and the preset historical data is input into the WT calculation module to extract the aging threshold. According to the relative size of the aging threshold and the aging warning index, a classified aging warning and model update strategy is constructed. The strategy includes: capacity decay warning, equivalent circuit model deviation warning, battery thermal model parameter warning and model normal.

[0012] Furthermore, before step S1, the method further includes:

[0013] First, perform OCV-SOC curve fitting:

[0014] In the cloud database, select the experimental data of the battery being fully charged under rated conditions, fully resting, and then intermittently discharging and resting, and fit the OCV-SOC curve with the polynomial shown in formula (1):

[0015]

[0016] Among them, V OC is the open circuit voltage (OCV) of the battery; z is the SOC of the battery; n is the order of the fitting polynomial, c i is the fitting coefficient;

[0017] Equivalent circuit model establishment:

[0018] The equivalent circuit model of the battery is established, and the time domain equation shown in formula (2) is established according to Kirchhoff's law:

[0019]

[0020] Where t is time, I is the load current, and the corresponding I(t) is the load current at time t. V p is the polarization voltage, V t is the terminal voltage, η is the coulomb efficiency of the battery, Q is the rated capacity of the battery, R s , R p , C p is the model parameter to be obtained, R s is the ohmic internal resistance, Rp is the polarization resistance, C p is the polarized capacitance;

[0021] Select input as I L , output y k =V t,k , the state vector x k =[V p,k z k ] T , transform the above equation into a state space equation and discretize it, as shown in equation (3), where Δt is the sampling step length;

[0022] x k =Ax k-1 +BI k-1 +w k

[0023] y k =f(z k )+V p,k +R s I k +v k (3)

[0024] Where: w k represents random process noise and disturbance, v k represents the interference of measurement noise and transmission noise;

[0025]

[0026] Thermal model establishment:

[0027] The mathematical expression of the battery thermal model is as follows:

[0028]

[0029] Among them, T s , T c , T f , respectively represent the surface temperature, internal temperature and ambient temperature of the battery, C s , C c , respectively represent the surface heat capacity and internal heat capacity of the battery, R c ,R u , represent the external thermal resistance and internal thermal resistance respectively. As the battery is continuously charged and discharged under operating conditions, the thermal model is used to estimate the internal heat generation of the battery caused by the current, and then the internal temperature and surface temperature of the battery are calculated;

[0030] Finally, the model parameters are identified and queried:

[0031] Using the HPPC test condition data of the battery of this model at different temperatures and different aging degrees in the cloud database, the least squares LS algorithm is used to identify the model parameters of the first-order RC equivalent circuit model mentioned above, and the mapping relationship between the equivalent circuit model parameters and the battery SOC, temperature and aging degree is obtained. At the same time, according to the heat generation and material properties of the charging and discharging process, the mapping relationship between the thermal model parameters and the battery aging degree and temperature is calibrated, and the above content is stored in the cloud database as a table to be queried;

[0032] [R s R p C p ]=mapping(T,SOC,age) (5)

[0033] [R c R u C c C s ]=mapping(T s ,age)

[0034] According to the corresponding mapping relationship, the corresponding first-order RC equivalent circuit model parameter values ​​and battery thermal model parameter values ​​can be obtained according to a specific set of SOC, temperature values ​​and aging degrees, which are used as known parameter values ​​for subsequent use of the model to estimate the battery state.

[0035] Further, the step S1 comprises:

[0036] Based on the equivalent circuit model established by formula (2), an observer based on the Kalman filter is established after the initial value is given. Through the following iterative process, the optimal state estimation based on the minimum root mean square error is performed on the state variables and output variables of each step to achieve noise filtering;

[0037] Battery status estimate update:

[0038] Error covariance estimate update:

[0039] Kalman gain matrix:

[0040] State estimation observation update:

[0041] Error covariance observation update:

[0042] In the above formula: P k represents the estimation error covariance matrix, Q k represents the random process noise w k The covariance matrix, R k represents the measurement noise vk The covariance matrix of the state value can be obtained by iterating the above steps. It can eliminate the influence of random noise and measurement noise on data.

[0043] Further, in step S3, extracting the aging threshold WT includes:

[0044] In order to evaluate the initial model error introduced in the model identification stage and characterize the identification error of the cloud battery model, the reasonable aging threshold WT of the model is extracted through the error level in the model parameter identification stage. The calculation formula is as follows:

[0045]

[0046] Where: X represents the state quantity, voltage, SOC, temperature; such as WT SOC That is, it represents the aging threshold calculated by SOC value; k represents the segment index. The parameters are considered to be the same within a certain SOC interval. Therefore, the mean absolute percentage error MAPE between the model estimation value and the theoretical value in each SOC interval is calculated respectively, and the maximum value is taken as the basis of the aging threshold calculation formula; α X Represents the adjustment factor. The accuracy requirements of the battery model in different working environments are different, and thus the tolerance range for the degree of aging is also different. By changing the size of the adjustment factor, further fine-tuning is performed based on the baseline value.

[0047] Furthermore, in step S3, the design of the aging warning indicator module includes:

[0048] In order to decouple part of the deviation caused by battery aging from the error between the model estimation value and the measured value, an aging warning indicator is designed as shown in the following formula:

[0049]

[0050] Where: X represents the state quantity, voltage, SOC and temperature, such as AWI SOC That is, it represents the aging warning index calculated by SOC value; since the received data contains loss interference, the state value of the corresponding part cannot be estimated by the model after the current data is lost, and the model will calculate the estimated value from the first bit of each received data; in the design of aging warning index, error analysis is directly performed on all received observation data, and the mean absolute percentage error MAPE between the model estimated value and the measured value of all observation part data is calculated as the aging warning index;

[0051] If the battery has not aged, the calculated aging warning index is less than the designed aging threshold. As the battery ages, the aging warning index gradually increases until it exceeds the aging warning index, thereby triggering the corresponding battery aging warning threshold.

[0052] Furthermore, according to the relative size of the aging threshold WT and the aging warning index AWI, the following classification aging warning and model update strategy is designed;

[0053] Capacity decline warning:

[0054] When the aging warning indicator calculated by the SOC value exceeds its aging warning threshold, the capacity decline warning is triggered:

[0055] AWI SOC ≥WT SOC ;

[0056] The SOC value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the capacity decays, the SOC value estimation error will increase directly, and eventually the SOC aging warning index exceeds the corresponding aging threshold. At this time, the capacity decay warning is triggered. The battery capacity needs to be recalibrated according to the most recent historical data and the SOC data needs to be estimated again until the aging warning index meets the threshold requirement.

[0057] Equivalent circuit model deviation warning:

[0058] When the aging warning index calculated by the voltage value exceeds its aging warning threshold, the equivalent circuit model deviation warning is triggered:

[0059] AWI V ≥WT V ;

[0060] The voltage value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the equivalent circuit model parameters deviate, the estimated error of the voltage value will increase directly, and eventually the aging warning index of the voltage value will exceed the corresponding aging threshold. At this time, the equivalent circuit model deviation warning is triggered. It is necessary to re-identify the battery equivalent circuit model based on the most recent historical data and estimate the voltage data again until the aging warning index meets the threshold requirement.

[0061] Battery thermal model parameter warning:

[0062] When the aging warning indicator calculated by the temperature value exceeds its aging warning threshold, the battery thermal model parameter warning is triggered:

[0063] AWI T ≥WT T ;

[0064] The voltage value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the thermal model parameters deviate, the estimated error of the voltage value will increase directly, and eventually the temperature aging warning index will exceed the corresponding aging threshold. At this time, the thermal model deviation warning is triggered. It is necessary to re-identify the thermal model parameters based on the most recent historical data and estimate the temperature data again until the aging warning index meets the threshold requirements.

[0065] Model is normal:

[0066] When all aging warning indicators are within the aging threshold range, it is considered that the aging degree is very slight and can be ignored, the model accuracy is within a reasonable range, and the current model can be used for subsequent battery management algorithm applications;

[0067]

[0068] Compared with the prior art, the advantages of the present invention are:

[0069] 1. The battery aging classification and early warning system applied to the "cloud-edge" battery management system can evaluate the degree of battery aging when the cloud BMS receives lost data. According to the designed aging indicators and aging threshold methods, it can decouple the aging caused by different parameters and classify the aging types, filling the gap in related fields and providing a practical solution for the aging classification and early warning of CEBMS.

[0070] 2. The aging classification warning method applied to the cloud battery management system in the "cloud-edge" scenario can provide reference and guidance for the parameter calibration and model update of the cloud battery model in CEBMS, and guide the update of specific model parameters in a more targeted manner, thereby ensuring the accuracy of the battery model established in the cloud BMS and improving the stability and accuracy of CEBMS. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 , Cloud Edge Battery Management System (CEBMS) structure diagram;

[0072] Figure 2 , structural diagram of the aging warning and classification module of the present invention;

[0073] Figure 3 , Comparison chart of fitted OCV-SOC curve and measured value;

[0074] Figure 4 , one 1-order RC equivalent circuit model diagram;

[0075] Figure 5 , battery thermal model diagram;

[0076] Figure 6, aging warning and classification verification result diagram. DETAILED DESCRIPTION

[0077] The specific implementation mode of the present invention is described below in conjunction with embodiments:

[0078] It should be noted that the structures, proportions, sizes, etc. shown in this specification are only used to match the contents disclosed in the specification so that people familiar with this technology can understand and read them, and are not used to limit the conditions under which the present invention can be implemented. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0079] At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" cited in this specification are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments to their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0080] Definitions of Abbreviations and Key Terms:

[0081] Battery Management System (BMS)

[0082] Cloud-Edge Battery Management System (CEBMS)

[0083] Open circuit voltage (OCV)

[0084] Battery State of Charge (SOC)

[0085] Aging Warning Indicator (AWI)

[0086] Warning Threshold (WT)

[0087] Embodiment 1:

[0088] Based on the shortcomings of the prior art, in order to fill the technical gap of the battery aging classification warning method in the scenario of data loss in CEBMS, the present invention designs a method that can classify and warn the aging of the actual battery system based on the battery model in the cloud BMS in the scenario of simultaneous loss of multi-dimensional data. The purpose of the present invention is achieved through the following technical solutions:

[0089] First, we established Figure 1 The CEBMS structure shown in the figure. The cloud battery management system mainly consists of three parts: vehicle side, cloud side and communication network. The working process is as follows:

[0090] Step 1. Vehicle-side data collection and transmission

[0091] The vehicle-side BMS uses the onboard sampling module to collect data such as the single-cell terminal voltage, total current, and single-cell surface temperature of the vehicle power battery system in each sampling cycle, and then encodes and packages the sampled data according to the preset communication protocol and sends it to the vehicle-side communication module. The communication module sends the original sampled data to the cloud system through far-field wireless communication technology;

[0092] Step 2. Cloud BMS aging classification warning module

[0093] The cloud BMS uses the communication module to receive the data sent by the vehicle and decodes it to obtain the original data, and inputs the obtained original data into the aging warning module designed by the present invention, and performs battery aging classification warning based on the received data containing noise and loss interference. If the aging warning is triggered, it is necessary to update the corresponding parameters of the cloud battery model according to the type of warning and evaluate the updated model again until it meets the accuracy requirements and then use the updated model as the latest model for subsequent applications.

[0094] Step 3. Battery management and data storage

[0095] The cloud BMS uses the battery model that meets the accuracy requirements and the received data as the input of the advanced battery management algorithm, and outputs the result of the battery state estimation, which will be sent to the communication module. In the communication module, the state estimation result will be encoded and packaged according to the preset communication protocol and sent to the vehicle-side system. In addition, the data will also be stored in the cloud database as a large battery data set, which can be used as offline data;

[0096] Step 4. Vehicle-side data reception and application

[0097] The vehicle-side BMS uses the communication module to receive data sent from the cloud and decode it to obtain the battery status estimation result, and then controls the vehicle-mounted battery system accordingly based on the estimated value.

[0098] In the structure of the cloud-edge battery management system, the design of the aging classification warning module in the cloud BMS is the core content of the present invention, which mainly includes three functions: threshold design, indicator design and judgment strategy. The overall structure is as follows Figure 2As shown, AWI (Aging Warning Indicator, AWI) is the designed aging warning indicator, and WT (Warning Threshold) is the aging threshold involved. The design method of the aging classification warning module is described in detail below.

[0099] 1 Model building and parameter identification

[0100] 1.1 OCV-SOC curve fitting

[0101] In the cloud database, select the experimental data of this type of battery under rated working conditions, fully charged, fully rested, and then intermittently discharged and rested. Since factors such as temperature and aging have little effect on OCV in a short time scale, the influence of related factors is ignored. The polynomial shown in formula (1) is used to fit the OCV-SOC curve.

[0102]

[0103] Among them, V OC is the open circuit voltage (OCV) of the battery; z is the SOC of the battery; n is the order of the fitting polynomial, and in the embodiment of the present application, n=5, c i is the fitting coefficient, and the fitted OCV-SOC curve is compared with the measured value. Figure 3 shown.

[0104] 1.2 Battery Modeling

[0105] 1.2.1 Establishment of equivalent circuit model

[0106] Establish the equivalent circuit model of the battery, in this application example, as follows Figure 4 The first-order RC model shown in the figure is used, and the time domain equation shown in equation (2) is established according to Kirchhoff's law.

[0107]

[0108] Where t is time, I is the load current, and the corresponding I(t) is the load current at time t. V p is the polarization voltage, V t is the terminal voltage, η is the coulomb efficiency of the battery, Q is the rated capacity of the battery, R s , R p , C p is the model parameter to be obtained, R s is the ohmic internal resistance, R p is the polarization resistance, C p is a polarized capacitor.

[0109] Select input as I L , output y k =Vt,k , the state vector x k =[V p,k z k ] T , transform the above equation into a state space equation and discretize it, as shown in equation (3). Where Δt is the sampling step length, and in this example, Δt = 1s:

[0110] x k =Ax k-1 +BI k-1 +w k

[0111] y k =f(z k )+V p,k +R s I k +v k (3)

[0112] Where: w k represents random process noise and disturbance, v k Represents the interference of measurement noise and transmission noise.

[0113]

[0114] 1.2.2 Thermal model establishment

[0115] To estimate the surface temperature and internal temperature of the battery, a battery thermal model needs to be established. The thermal conduction of lithium-ion batteries mainly depends on the radial thermal resistance value, ignoring the axial thermal resistance. Figure 5 Battery thermal model shown:

[0116] The mathematical expression of the above battery thermal model is as follows:

[0117]

[0118] Among them, T s , T c , T f , respectively represent the surface temperature, internal temperature and ambient temperature of the battery, C s , C c , respectively represent the surface heat capacity and internal heat capacity of the battery, R c ,R u , represent the external thermal resistance and internal thermal resistance respectively. As the battery is continuously charged and discharged under operating conditions, the thermal model can be used to estimate the internal heat generation of the battery caused by the current, and then calculate the internal temperature and surface temperature of the battery.

[0119] 1.3 Model parameter identification and query

[0120] Using the HPPC test condition data of this type of battery at different temperatures and different aging degrees in the cloud database, the least squares (LS) algorithm is used to identify the model parameters of the aforementioned first-order RC equivalent circuit model, and the mapping relationship between the equivalent circuit model parameters and the battery SOC, temperature and aging degree is obtained. At the same time, according to the heat generation and material properties of the charging and discharging process, the mapping relationship between the thermal model parameters and the battery aging degree and temperature is calibrated, and the above content is stored in the cloud database as a table to be queried.

[0121] [R s R p C p ]=mapping(T,SOC,age) (5)

[0122] [R c R u C c C s ]=mapping(T s ,age)

[0123] According to the corresponding mapping relationship, the corresponding first-order RC equivalent circuit model parameter values ​​and battery thermal model parameter values ​​can be obtained according to a specific set of SOC, temperature values ​​and aging degrees, as the known parameter values ​​for the subsequent use of the model to estimate the battery state. In addition, the cloud battery management system uses the large-scale computing power resources of its distributed cloud computing and the massive battery data in the cloud database, combined with the battery aging warning method proposed in this design, and when the warning is triggered, the newly generated battery data set is used to iteratively update the model parameter table to fully ensure the accuracy of the battery model.

[0124] 2.1 Data cleaning and denoising based on Kalman filter

[0125] In the CEBMS system, the vehicle-side BMS, as an edge node, sends data to the cloud BMS through far-field wireless communication technology. The data collection and transmission process is often affected by measurement noise and electromagnetic interference. In order to reduce the impact of measurement noise on the battery aging warning results, it is necessary to first filter the received data for noise.

[0126] Based on the equivalent circuit model established by formula (2), an observer based on the Kalman filter is established after the initial value is given. Through the following iterative process, the optimal state estimation based on the minimum root mean square error is performed on the state variables and output variables of each step to achieve noise filtering.

[0127] 1. Battery status estimation update:

[0128] 2. Error covariance estimate update:

[0129] 3. Kalman gain matrix:

[0130] 4. State estimation observation update:

[0131] 5. Error covariance observation update:

[0132] In the above formula: P k represents the estimation error covariance matrix, Q k represents the random process noise w k The covariance matrix, R k represents the measurement noise v k The covariance matrix of . By iterating the above steps, we can get the optimal estimate of the state value. It can eliminate the influence of random noise and measurement noise on data. In addition, the Kalman filter algorithm can make the state value gradually converge to the true value with iterative operation, without being affected by the initial value error, and eliminate the cumulative error.

[0133] 2.2 Aging Threshold (WT) Extraction

[0134] The original model error will be introduced in the parameter identification process. As the battery pack charge and discharge cycles increase, the aging of the battery system will also cause a gradual deviation between the cloud battery model and the actual battery. Both deviations will appear as errors between the model estimate and the measured value. The aging threshold is designed to evaluate the initial model error introduced in the model identification stage. In order to characterize the identification error of the cloud battery model, the reasonable aging threshold (WT) of the model is extracted through the error level in the model parameter identification stage. The calculation formula is as follows:

[0135]

[0136] Where: X represents the state quantity (voltage, SOC, temperature), etc., such as WT SOC That is, it represents the aging threshold calculated by the SOC value; k represents the segment index, which is similar to the common battery equivalent circuit model parameter identification. The parameters are considered to be the same within a certain SOC interval (the SOC interval is taken as 5% in the embodiment of the present application). Therefore, the mean absolute percentage error (MAPE) between the model estimation value and the theoretical value in each SOC interval is calculated respectively, and the maximum value thereof is taken as the basis for the aging threshold calculation formula; α X represents the adjustment factor. The accuracy requirements of the battery model under different working environments are different, and thus the tolerance range of the aging degree is also different. By changing the size of the adjustment factor, further fine-tuning can be performed on the basis of the reference value. In order to make the aging threshold design more in line with the needs of actual scenarios, α is taken in the embodiment of the present application. SOC =0.8,αV =2,α T =1.

[0137] The design of the aging threshold introduces the innovative point of segmented extraction error, which can comprehensively and accurately evaluate the adaptability and error level of each model parameter and fully evaluate the model identification error. At the same time, the segmented calculation error can also eliminate the influence of cumulative error, avoiding the excessive threshold design due to cumulative error, which in turn causes a large missed diagnosis rate. In addition, the introduction of adjustment factors also provides more adaptability for threshold design.

[0138] 2.3 Design of Aging Warning Index (AWI)

[0139] In order to decouple the partial deviation caused by battery aging from the error between the model estimated value and the measured value, the present invention designs an aging warning indicator as shown in the following formula:

[0140]

[0141] Where: X represents the state quantity (voltage, SOC, temperature), etc., such as AWI SOC That is, it represents the aging warning indicator calculated by SOC value; because the received data contains loss interference, the state value of the corresponding part cannot be estimated by the model after the current data is lost. The model will calculate the estimated value from the first bit of each received data, thus eliminating the cumulative error. In order to analyze the information of the observed data to the greatest extent, the aging warning indicator design directly performs error analysis on all received observation data, and calculates the mean absolute percentage error (MAPE) between the model estimated value and the measured value of all observed data as the aging warning indicator.

[0142] If the battery has not aged, the calculated aging warning index should be less than the aging threshold designed in 2.1.2. As the battery ages, the aging warning index gradually increases until it exceeds the aging warning index, thereby triggering the corresponding battery aging warning threshold.

[0143] 2.4 Classification warning and model updating strategy

[0144] According to the relative sizes of the aging threshold (WT) and the aging warning index (AWI) designed in the above section, the present invention designs the following classified aging warning and model update strategy.

[0145] a. Capacity decline warning

[0146] When the aging warning indicator calculated by the SOC value exceeds its aging warning threshold, the capacity decline warning is triggered:

[0147] AWI SOC ≥WT SOC

[0148] The error of SOC value is mainly affected by battery capacity and current value. In the loss scenario, current data is incomplete. Therefore, the SOC value is estimated based on the observed part of current data, and then the aging warning index is calculated. When the capacity declines, it will directly lead to an increase in the error of SOC value estimation, and eventually the aging warning index of SOC exceeds the corresponding aging threshold, triggering the capacity decline warning. It is necessary to recalibrate the battery capacity according to the most recent historical data and estimate the SOC data again until the aging warning index meets the threshold requirements.

[0149] b. Equivalent circuit model deviation warning

[0150] When the aging warning index calculated by the voltage value exceeds its aging warning threshold, the equivalent circuit model deviation warning is triggered:

[0151] AWI V ≥WT V

[0152] The error in the voltage value is mainly affected by the equivalent circuit model and the current value. In the loss scenario, the current data is incomplete. Therefore, the voltage value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the equivalent circuit model parameters deviate, the estimated error of the voltage value will directly increase, and eventually the aging warning index of the voltage value will exceed the corresponding aging threshold. At this time, the equivalent circuit model deviation warning is triggered. It is necessary to re-identify the battery equivalent circuit model based on the most recent historical data and estimate the voltage data again until the aging warning index meets the threshold requirements.

[0153] c. Battery thermal model parameter warning

[0154] When the aging warning indicator calculated by the temperature value exceeds its aging warning threshold, the battery thermal model parameter warning is triggered:

[0155] AWI T ≥WT T

[0156] The error of temperature value is mainly affected by the battery thermal model and current value. In the loss scenario, the current data is incomplete. Therefore, the voltage value is estimated based on the observed part of the current data, and then the aging warning index is calculated. When the thermal model parameters deviate, it will directly lead to an increase in the estimated error of the voltage value, and eventually the temperature aging warning index exceeds the corresponding aging threshold. At this time, the thermal model deviation warning is triggered. It is necessary to re-identify the thermal model parameters based on the most recent historical data and estimate the temperature data again until the aging warning index meets the threshold requirements.

[0157] d. Model is normal

[0158] When all aging warning indicators are within the aging threshold range, it is considered that the degree of aging is very slight and can be ignored, the model accuracy is within a reasonable range, and the current model can be used for subsequent battery management algorithm applications.

[0159]

[0160] Embodiment 2:

[0161] The scheme designed according to the above content can realize the evaluation of model accuracy and battery system aging level in CEBMS under data loss scenario, and realize battery aging classification warning under data loss scenario under CEBMS architecture. The following content verifies the scheme designed by the present invention through experiments and simulations.

[0162] By increasing the DC internal resistance R in the equivalent circuit model s And reduce the battery capacity Q to simulate the aging of the battery system. And calculate the aging warning indicators under different aging degrees to obtain Figure 6 The warning results shown are:

[0163] From the results in the figure, it can be seen that the aging classification warning method designed by the present invention increases the voltage warning index value quickly while the SOC warning index value changes little during the process of increasing DC internal resistance; correspondingly, the SOC warning index value increases quickly while the voltage warning index value changes little during the process of capacity decay. The results show that this method can effectively detect the occurrence of battery aging and can better classify the type of aging, providing effective guidance for the parameter update of the model.

[0164] According to the above steps, the accuracy of the cloud battery model and the aging level of the battery system can be evaluated and the aging types can be classified based on the missing data received by the cloud BMS platform, realizing the battery aging classification warning function under the CEBMS architecture and further providing a cloud model update strategy, which provides a reliable method to ensure the accuracy of the cloud battery model and can serve as a strong basis for guiding its parameter updates.

[0165] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

[0166] Many other changes and modifications may be made without departing from the concept and scope of the present invention.It should be understood that the present invention is not limited to the specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. A battery aging classification warning method for CEBMS in data loss scenarios, It is characterized in that The method comprises: S1: De-noise the original data based on the Kalman filter to obtain the de-noised voltage, current, SOC and temperature data; S2: Input the current data into the electrical model and the thermal model to obtain the estimated voltage, SOC and temperature data; S3: The estimated data is input into the AWI calculation module, and the preset historical data is input into the WT calculation module to extract the aging threshold. According to the relative size of the aging threshold and the aging warning index, a classified aging warning and model update strategy is constructed. The strategy includes: capacity decay warning, equivalent circuit model deviation warning, battery thermal model parameter warning and model normal.

2. According to claim 1, a battery aging classification warning method for CEBMS in a data loss scenario, It is characterized in that Before step S1, the method further includes: First, perform OCV-SOC curve fitting: In the cloud database, select the experimental data of the battery being fully charged under rated conditions, fully resting, and then intermittently discharging and resting, and fit the OCV-SOC curve with the polynomial shown in formula (1): Among them, V OC is the open circuit voltage (OCV) of the battery; z is the SOC of the battery; n is the order of the fitting polynomial, c i is the fitting coefficient; Equivalent circuit model establishment: The equivalent circuit model of the battery is established, and the time domain equation shown in formula (2) is established according to Kirchhoff's law: Where t is time, I is the load current, and the corresponding I(t) is the load current at time t. V p is the polarization voltage, V t is the terminal voltage, η is the coulomb efficiency of the battery, Q is the rated capacity of the battery, R s , R p , C p is the model parameter to be obtained, R s is the ohmic internal resistance, R p is the polarization resistance, C p is the polarized capacitance; Select input as I L , output y k =V t,k , the state vector x k =[V p,k z k ] T , transform the above equation into a state space equation and discretize it, as shown in equation (3), where Δt is the sampling step length; x k =Ax k-1 +BI k-1 +w k y k =f(z k )+V p,k +R s I k +v k (3) Where: w k represents random process noise and disturbance, v k represents the interference of measurement noise and transmission noise; Thermal model establishment: The mathematical expression of the battery thermal model is as follows: Among them, T s , T c , T f , respectively represent the surface temperature, internal temperature and ambient temperature of the battery, C s , C c , respectively represent the surface heat capacity and internal heat capacity of the battery, R c ,R u , represent the external thermal resistance and internal thermal resistance respectively. As the battery is continuously charged and discharged under operating conditions, the thermal model is used to estimate the internal heat generation of the battery caused by the current, and then the internal temperature and surface temperature of the battery are calculated; Finally, the model parameters are identified and queried: Using the HPPC test condition data of the battery of this model at different temperatures and different aging degrees in the cloud database, the least squares LS algorithm is used to identify the model parameters of the first-order RC equivalent circuit model mentioned above, and the mapping relationship between the equivalent circuit model parameters and the battery SOC, temperature and aging degree is obtained. At the same time, according to the heat generation and material properties of the charging and discharging process, the mapping relationship between the thermal model parameters and the battery aging degree and temperature is calibrated, and the above content is stored in the cloud database as a table to be queried; [R s R p C p ]=mapping(T,SOC,age) (5) [R c R u C c C s ]=mapping(T s ,age) According to the corresponding mapping relationship, the corresponding first-order RC equivalent circuit model parameter values ​​and battery thermal model parameter values ​​can be obtained according to a specific set of SOC, temperature values ​​and aging degrees, which are used as known parameter values ​​for subsequent use of the model to estimate the battery state.

3. According to claim 1, a battery aging classification warning method for CEBMS in a data loss scenario, It is characterized in that The step S1 comprises: Based on the equivalent circuit model established by formula (2), an observer based on the Kalman filter is established after the initial value is given. Through the following iterative process, the optimal state estimation based on the minimum root mean square error is performed on the state variables and output variables of each step to achieve noise filtering; Battery status estimate update: Error covariance estimate update: Kalman gain matrix: State estimation observation update: Error covariance observation update: In the above formula: P k represents the estimation error covariance matrix, Q k represents the random process noise w k The covariance matrix, R k represents the measurement noise v k The covariance matrix of the state value can be obtained by iterating the above steps. It can eliminate the influence of random noise and measurement noise on data.

4. According to claim 1, a battery aging classification warning method for CEBMS in a data loss scenario, It is characterized in that In step S3, extracting the aging threshold WT includes: In order to evaluate the initial model error introduced in the model identification stage and characterize the identification error of the cloud battery model, the reasonable aging threshold WT of the model is extracted through the error level in the model parameter identification stage. The calculation formula is as follows: Where: X represents the state quantity, voltage, SOC, temperature; such as WT SOC That is, it represents the aging threshold calculated by SOC value; k represents the segment index. The parameters are considered to be the same within a certain SOC interval. Therefore, the mean absolute percentage error MAPE between the model estimation value and the theoretical value in each SOC interval is calculated respectively, and the maximum value is taken as the basis of the aging threshold calculation formula; α X Represents the adjustment factor. The accuracy requirements of the battery model in different working environments are different, and thus the tolerance range for the degree of aging is also different. By changing the size of the adjustment factor, further fine-tuning is performed based on the baseline value.

5. According to claim 1, a battery aging classification warning method for CEBMS in a data loss scenario, It is characterized in that In step S3, the design of the aging warning indicator module includes: In order to decouple part of the deviation caused by battery aging from the error between the model estimation value and the measured value, an aging warning indicator is designed as shown in the following formula: Where: X represents the state quantity, voltage, SOC and temperature, such as AWI SOC That is, it represents the aging warning index calculated by SOC value; since the received data contains loss interference, the state value of the corresponding part cannot be estimated by the model after the current data is lost, and the model will calculate the estimated value from the first bit of each received data; in the design of aging warning index, error analysis is directly performed on all received observation data, and the mean absolute percentage error MAPE between the model estimated value and the measured value of all observation part data is calculated as the aging warning index; If the battery has not aged, the calculated aging warning index is less than the designed aging threshold. As the battery ages, the aging warning index gradually increases until it exceeds the aging warning index, thereby triggering the corresponding battery aging warning threshold.

6. According to claim 1, a battery aging classification warning method for CEBMS in a data loss scenario, It is characterized in that According to the relative size of the aging threshold WT and the aging warning index AWI, the following classification aging warning and model update strategy is designed; Capacity decline warning: When the aging warning indicator calculated by the SOC value exceeds its aging warning threshold, the capacity decline warning is triggered: AWI SOC ≥WT SOC ; The SOC value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the capacity decays, the SOC value estimation error will increase directly, and eventually the SOC aging warning index exceeds the corresponding aging threshold. At this time, the capacity decay warning is triggered. The battery capacity needs to be recalibrated according to the most recent historical data and the SOC data needs to be estimated again until the aging warning index meets the threshold requirement. Equivalent circuit model deviation warning: When the aging warning index calculated by the voltage value exceeds its aging warning threshold, the equivalent circuit model deviation warning is triggered: AWI V ≥WT V ; The voltage value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the equivalent circuit model parameters deviate, the estimated error of the voltage value will increase directly, and eventually the aging warning index of the voltage value will exceed the corresponding aging threshold. At this time, the equivalent circuit model deviation warning is triggered. It is necessary to re-identify the battery equivalent circuit model based on the most recent historical data and estimate the voltage data again until the aging warning index meets the threshold requirement. Battery thermal model parameter warning: When the aging warning indicator calculated by the temperature value exceeds its aging warning threshold, the battery thermal model parameter warning is triggered: AWI T ≥WT T ; The voltage value is estimated based on the observed partial current data, and then the aging warning index is calculated. When the thermal model parameters deviate, the estimated error of the voltage value will increase directly, and eventually the temperature aging warning index will exceed the corresponding aging threshold. At this time, the thermal model deviation warning is triggered. It is necessary to re-identify the thermal model parameters based on the most recent historical data and estimate the temperature data again until the aging warning index meets the threshold requirements. Model is normal: When all aging warning indicators are within the aging threshold range, it is considered that the aging degree is very slight and can be ignored, the model accuracy is within a reasonable range, and the current model can be used for subsequent battery management algorithm applications;

Citation Information

Cited By

  • A batch upgrade test device and test method for distributed equipment

    CN119739641A