Vehicle cloud collaborative SOH estimation method for real vehicle operation data

By using hybrid data-driven modeling and statistical feature indexes of SOH and ΔSOH in vehicle-cloud collaboration, the SOH estimation problem in the existing technology under non-constant current and constant voltage conditions is solved, and the real world is achieved Accurate and stable SOH estimation ensures battery safety and reliable operation of electric vehicles.

CN118938007APending Publication Date: 2024-11-12HARBIN INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410921981.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-10
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing vehicle-cloud collaborative SOH estimation method has problems in aging feature structure and heterogeneous data-driven modeling under non-constant current and constant voltage conditions, and verification based on laboratory ideal cycle aging data does not conform to the real-world driving environment.

Method used

A hybrid data-driven modeling strategy of SOH and ΔSOH is adopted to construct aging characteristics based on statistical feature indicators, and SOH is estimated in vehicle-cloud collaboration to adapt to the actual operating environment.

Benefits of technology

Provide stable SOH estimation results under real-world conditions, taking into account the impact of multi-stage constant current charging and heterogeneous data, ensuring safe use of batteries and safe driving of electric vehicles.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938007A_ABST
    Figure CN118938007A_ABST
Patent Text Reader

Abstract

The invention discloses a vehicle cloud collaborative SOH estimation method for real vehicle operation data, and relates to the technical field of lithium ion batteries. A hybrid data driving estimation model based on SOH and delta SOH is established at a vehicle end and a cloud end, updating of the SOH data driving estimation model is performed by monitoring a full-charging stage, an SOH estimation result obtained after updating is used for updating the delta SOH data driving estimation model, and an obtained SOH estimation value is used as an SOH estimation result of a non-full-charging stage; selecting statistical characteristic indexes to construct aging characteristics, selecting parameters for constructing the aging characteristics, calculating the aging characteristics, and performing online incremental calculation on the aging characteristics; and designing self-monitoring and model updating of the vehicle cloud end, mutual monitoring of the vehicle cloud end and a cooperative SOH estimation strategy. A modeling strategy is driven by adopting mixed data of SOH and delta SOH, aging characteristics are constructed based on statistical characteristic indexes, SOH estimation is performed by utilizing vehicle cloud cooperation, and the method is more in line with an actual working condition operation environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion batteries, and in particular to a vehicle-cloud collaborative SOH estimation method for real vehicle operation data. Background Art

[0002] With the development of technologies such as artificial intelligence, big data, 5G communications, and the Internet of Vehicles, the operating data collected by the vehicle side is uploaded to the cloud in real time. The higher computing power and stronger storage capacity of the cloud make it possible to establish an accurate and robust SOH estimation model, and the more accurate SOH estimation results obtained in the cloud can also monitor the SOH estimation results on the vehicle side. Conversely, the vehicle side can also provide accurate and necessary SOC estimation reference information for the cloud to establish an effective SOH estimation model based on more refined data collection and faster data processing and response capabilities. Therefore, the SOH estimation strategy of vehicle-cloud collaboration can obtain more stable and reliable SOH estimation results, and also makes the state estimation based on vehicle-cloud collaboration gradually become a research hotspot in the field of battery state estimation.

[0003] However, the SOH estimation methods for vehicle-cloud collaboration proposed in existing work are still traditional SOH estimation methods. Most of these methods establish deep learning models on the cloud side and aging experience models on the vehicle side, and then use the Kalman filter algorithm to fuse the SOH estimation results obtained on the vehicle side and the cloud side. Problems in traditional SOH data-driven estimation, such as aging feature construction under non-constant current and constant voltage conditions, and SOH data-driven modeling based on heterogeneous data, still exist. In addition, most of the existing methods are verified based on ideal laboratory cycle aging data (such as NASA's public aging data set), which obviously does not conform to the actual driving environment in the real world. Therefore, the research on vehicle-cloud collaborative SOH estimation should be oriented towards the actual operating environment of the real world, and it is necessary to consider issues such as aging feature construction under complex charging modes and random operating conditions, data-driven modeling under conditions such as heterogeneous data or limited data-driven modeling data. Summary of the invention

[0004] In order to address the shortcomings of the background technology, the present invention provides a vehicle-cloud collaborative SOH estimation method for real vehicle operation data. It adopts a hybrid data-driven modeling strategy of SOH and ΔSOH, constructs aging characteristics based on statistical characteristic indicators, and uses vehicle-cloud collaboration to perform SOH estimation, which is more in line with the actual operating environment.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a vehicle-cloud collaborative SOH estimation method for real vehicle operation data, comprising the following steps:

[0006] Step 1: Hybrid data-driven modeling of SOH and ΔSOH

[0007] The SOH data-driven estimation model is characterized as follows:

[0008] Μ SOH =f(Ψ SOH ,SOH) (1)

[0009] Where f represents the data-driven estimation algorithm used to establish the SOH data-driven estimation model, Ψ SOH Represents the constructed modeling features for SOH data-driven;

[0010] Taking the difference value ΔSOH of SOH as the modeling target, the ΔSOH data-driven estimation model is characterized as follows:

[0011] Μ ΔSOH =f'(Ψ ΔSOH ,ΔSOH) (2)

[0012] Where f' represents the data-driven estimation algorithm used to establish the ΔSOH data-driven estimation model, Ψ ΔSOH represents the constructed modeling features for ΔSOH data-driven;

[0013] A hybrid data-driven estimation model based on SOH and ΔSOH is designed. The SOH data-driven estimation model and the ΔSOH data-driven estimation model are established on both the vehicle side and the cloud side. The characteristics are as follows:

[0014]

[0015] The SOH data-driven estimation model is updated by monitoring the full-charge stage, wherein the full-charge stage refers to the interval state of charge ΔSOC ≥ 80% during the charging process, and the SOH estimation result obtained by the updated SOH data-driven estimation model is used to update the SOH estimation value obtained by the ΔSOH data-driven estimation model, and the SOH estimation value obtained by the ΔSOH data-driven estimation model is used as the SOH estimation result in the non-full-charge stage;

[0016] Step 2: Construction of aging characteristics based on statistical characteristic indicators

[0017] S2.1. Aging feature construction for SOH data-driven modeling

[0018] a) Selection of statistical characteristic indicators for constructing aging characteristics

[0019] In the SOH data-driven modeling, in addition to selecting the standard deviation and coefficient of variation representing the degree of data dispersion, and the skewness, kurtosis, and peak-to-skewness ratio representing the data distribution characteristics, the mean and quartile representing the degree of data dispersion, and the discrete entropy representing the degree of data chaos are also selected. A total of eight basic statistical characteristic indicators are selected to construct the aging characteristics;

[0020] b) Parameter selection for constructing aging features

[0021] To ensure the validity of the calculated aging characteristic value, the current is input into the exponential function equation. The parameters used to construct the aging characteristic are as follows:

[0022]

[0023] Where U t Represents the battery terminal voltage, I t Indicates the battery current;

[0024] c) Specific calculation method of aging characteristics

[0025] According to the statistical characteristic index for constructing the aging characteristic selected in S2.1 a) and the parameters for constructing the aging characteristic selected in S2.1 b), the aging characteristic is calculated based on the accumulation of the actual vehicle operating condition data as follows:

[0026] l SOH,k =F ea,j (D a ),D a ={X i ,i=1,...,k} (5)

[0027] In the formula, l SOH,k represents the aging characteristic value calculated at the kth moment for SOH estimation, F ea,j represents the selected statistical feature index, j = 1, ..., M, M is the number of statistical feature indicators, D a Represents the actual vehicle operating condition data accumulated from the first moment to the current k moment;

[0028] d) Online incremental calculation of aging characteristics

[0029] The vehicle side constructs aging features based on the mean, standard deviation, coefficient of variation, skewness, kurtosis, and peak-to-skewness ratio with online incremental calculation function. The online incremental calculation equations of the six relevant aging features used for vehicle side SOH data-driven modeling are as follows:

[0030] ①Online incremental calculation equation of aging characteristics constructed based on the mean value:

[0031]

[0032] in,

[0033]

[0034] In the formula, Characterizes the mean index calculated at the nth moment, Represents the mean index calculated at the n+1th moment, Xn+1 is the actual vehicle operating condition data at the n+1th moment, and n is the number of data used for feature calculation from the 1st moment to the current moment;

[0035] ②Online incremental calculation equation of aging characteristics constructed based on standard deviation:

[0036]

[0037] in,

[0038]

[0039] In the formula, Characterizes the standard deviation index calculated at the nth moment, Characterizes the standard deviation index calculated at the n+1th moment;

[0040] ③Online incremental calculation equation of aging characteristics constructed based on difference coefficient:

[0041]

[0042] in,

[0043]

[0044] In the formula, Characterizes the difference coefficient index calculated at the nth moment, Characterizes the difference coefficient index calculated at the n+1th moment;

[0045] ④Online incremental calculation equation based on aging characteristics constructed by skewness:

[0046]

[0047] in,

[0048]

[0049] In the formula, Characterizes the skewness index calculated at the nth moment, Characterizes the skewness index calculated at the n+1th moment;

[0050] ⑤Online incremental calculation equation based on aging characteristics constructed by kurtosis:

[0051]

[0052] in,

[0053]

[0054] In the formula, Characterizes the kurtosis index calculated at the nth moment, Characterizes the kurtosis index calculated at the n+1th moment;

[0055] ⑥Online incremental calculation equation based on aging characteristics constructed by peak deviation ratio:

[0056]

[0057] In the formula, Characterizes the peak deviation ratio index calculated at the n+1th moment;

[0058] S2.2. Construction of aging features for ΔSOH data-driven modeling

[0059] a) Selection of statistical characteristic indicators for constructing aging characteristics

[0060] Use the statistical characteristic indicators selected in S2.1 a) to construct aging characteristics for ΔSOH data-driven modeling;

[0061] b) Parameter selection for constructing aging features

[0062] The vehicle-side ΔSOH data-driven modeling selects the measurable parameters of the battery including voltage, current and temperature to construct the aging characteristics. The cloud-side ΔSOH data-driven modeling selects the measurable parameters of the battery including voltage, current and temperature and the measurable parameters of the vehicle including vehicle speed, longitude of vehicle location, latitude of vehicle location, accelerator pedal travel value and brake pedal travel value to construct the aging characteristics.

[0063] c) Specific calculation method of aging characteristics

[0064] For ΔSOH data-driven modeling, the aging characteristics are calculated based on the actual vehicle operating condition data with fixed mileage as follows:

[0065] l ΔSOH,k =F ea,j (D' a ),D' a ={X i ,i=km,...,k} (17)

[0066] In the formula, l ΔSOH,k represents the aging characteristic value calculated at the kth moment for ΔSOH estimation, D' a It represents the actual vehicle operating condition data accumulated from the fixed mileage, i.e., from the km moment to the current k moment, and m represents the data window length;

[0067] For ΔSOH data-driven modeling, the vehicle mileage and the SOH reference value obtained based on the battery reference capacity are fitted based on the polynomial fitting equation, where the battery reference capacity is calculated based on the full charge stage of the battery as follows:

[0068]

[0069] In the formula, Cap ref represents the reference capacity of the battery, and ΔQ represents the charging capacity of the battery during the charging process;

[0070] Therefore, on the basis of obtaining the smooth attenuation curve of battery aging, the ΔSOH target modeling value with real physical meaning is obtained as follows:

[0071]

[0072] In the formula, diff represents differential calculation, Poly represents the polynomial fitting equation, and M mileage Indicates the vehicle mileage value, SOH ref It represents the SOH reference value obtained based on the battery reference capacity, which is calculated as follows:

[0073]

[0074] In the formula, Cap nor Indicates the nominal capacity of the battery when it is not aged;

[0075] d) Online incremental calculation of aging characteristics

[0076] Since the aging features constructed by ΔSOH data-driven modeling are the same as those constructed by SOH data-driven modeling, the online incremental calculation equations of the relevant aging features in formulas (6) to (16) are also used for the online incremental calculation of the aging features of ΔSOH data-driven modeling;

[0077] Step 3: Vehicle-Cloud Collaborative SOH Estimation

[0078] S3.1. Self-monitoring and model updating in the car-cloud

[0079] The effectiveness of the SOH data-driven estimation model and the ΔSOH data-driven estimation model established on the vehicle side and the cloud side is monitored according to the obtained SOH reference value. If the following conditions are met, the SOH data-driven estimation model and the ΔSOH data-driven estimation model need to be updated;

[0080] |SOH ref -SOH Cloud / SOH Vehicle |>δ (21)

[0081] In the formula, SOH Cloud Represents the estimated SOH value obtained by the cloud. Vehicle represents the SOH estimation value obtained by the vehicle end, and δ is the set threshold for abnormal SOH estimation result;

[0082] The vehicle and cloud also need to monitor the SOH estimation results obtained based on the established SOH data-driven estimation model. Compared with the SOH estimation results obtained based on the established ΔSOH data-driven estimation model If the following conditions are met, it means that the SOH estimation results obtained are significantly different between the vehicle side and the cloud side;

[0083]

[0084] In this case, it is necessary to determine the time interval between the current moment and the last model update. When the time interval is less than or equal to half a month, the SOH estimation result based on ΔSOH data-driven modeling can be used as the SOH estimation result at the current moment. On the contrary, if the time interval is greater than half a month, the average of the SOH estimation result based on SOH data-driven modeling and the SOH estimation result based on ΔSOH data-driven modeling is used as the SOH estimation result at that moment.

[0085] S3.2. Vehicle-cloud mutual monitoring and collaborative SOH estimation strategy

[0086] In addition to monitoring and model updating on the vehicle side and the cloud side, a mutual monitoring strategy between the vehicle side and the cloud side needs to be designed to further ensure the validity of the SOH estimation results obtained by the vehicle side and the cloud side.

[0087] First, when it is impossible to obtain an accurate SOH reference value by fully charging the battery, a full charging phase is required to self-monitor and update the models on the vehicle side and the cloud side, and compare the difference between the SOH estimation results obtained on the vehicle side and the SOH estimation results obtained on the cloud side with the set threshold, as shown in the following formula:

[0088] |SOH Cloud -SOH Vehicle |>δ (23)

[0089] If formula (23) is satisfied, it is necessary to determine whether the model established on the cloud side has been updated during the full charging stage. If it has been updated, the SOH data-driven estimation model on the vehicle side is updated according to the SOH estimation result obtained on the cloud side. Otherwise, the average value of the SOH estimation results obtained on the vehicle side and the cloud side is taken as the SOH estimation value at the current moment. If formula (23) is not satisfied, there is no need to start the model update on the vehicle side and the cloud side.

[0090] Furthermore, when selecting statistical characteristic indicators for constructing aging characteristics in a) of S2.1, on the basis of the eight basic statistical characteristic indicators, combined statistical characteristic indicators are added, including: standard deviation*kurtosis, standard deviation / kurtosis, standard deviation*ln(kurtosis), standard deviation / ln(kurtosis), ln(standard deviation)*kurtosis, ln(standard deviation) / kurtosis, standard deviation*peak-to-peak ratio, coefficient of difference*ln(peak-to-peak ratio) and coefficient of difference / ln(peak-to-peak ratio).

[0091] Furthermore, the threshold δ for abnormal SOH estimation result in S3.1 is set to 2% or 3%.

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] 1. The method of the present invention fully considers the real-world vehicle operating conditions, and takes into account the impact of factors such as multi-stage constant current charging and heterogeneous data on data-driven modeling, which is more in line with SOH estimation in practical applications;

[0094] 2. The aging feature constructed by the method of the present invention takes into account the online incremental calculation function, so that the constructed aging feature can be easily deployed on the vehicle side in practical applications;

[0095] 3. The method of the present invention adopts a hybrid data-driven modeling strategy of SOH and ΔSOH, which can ensure that a relatively stable SOH estimation result can be provided even without sufficient modeling data;

[0096] 4. The method of the present invention can obtain stable SOH estimation results, which is helpful to maintain the safe use of batteries and ensure the safe driving of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a flowchart of the present invention;

[0098] Figure 2 This is a schematic diagram of the strategy for vehicle-cloud collaborative SOH estimation proposed in the present invention;

[0099] Figure 3 is the SOH estimation result obtained based on SOH data-driven modeling in the embodiment;

[0100] Figure 4 It is the SOH estimation result obtained based on ΔSOH data-driven modeling in the embodiment. DETAILED DESCRIPTION

[0101] The technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0102] like Figure 1-2 As shown in the figure, a vehicle-cloud collaborative SOH estimation method for real vehicle operation data is proposed. Figure 1 As shown, the following steps are included:

[0103] Step 1: Hybrid data-driven modeling of SOH and ΔSOH

[0104] The traditional SOH data-driven estimation method directly uses SOH as the data-driven modeling target. The accuracy and robustness of its modeling depend on the quantity and quality of the modeling data. The SOH data-driven estimation model is characterized as follows:

[0105] Μ SOH =f(Ψ SOH ,SOH) (1)

[0106] Where f represents the data-driven estimation algorithm used to establish the SOH data-driven estimation model, Ψ SOH Represents the constructed modeling features for SOH data-driven.

[0107] The cloud has a large amount of modeling data under actual vehicle operating conditions, and in theory it is easy to establish a robust SOH data-driven estimation model. However, these modeling data may come from different types of batteries or vehicle models, which are multi-source heterogeneous data, or they may come from the same type of batteries or vehicle models, but are affected by different driving habits, different geographical locations and other factors, which are homogeneous heterogeneous data. Regardless of whether it is multi-source heterogeneous data or homogeneous heterogeneous data, there are differences in marginal probability distribution or even conditional probability distribution between heterogeneous data, which may cause the traditional SOH data-driven estimation method to have a large estimation deviation in the SOH estimation results at a certain moment. In addition, the data storage and computing capabilities of the vehicle side are limited, and it is difficult to establish a robust SOH data-driven estimation model like the cloud side. Therefore, the SOH data-driven estimation model deployed on the vehicle side may also have a large estimation deviation.

[0108] In order to reduce the problem of large SOH estimation bias in traditional SOH data-driven estimation methods due to factors such as differences in probability distribution between heterogeneous data, a ΔSOH data-driven estimation model is proposed, which is characterized as follows:

[0109] Μ ΔSOH =f'(Ψ ΔSOH ,ΔSOH) (2)

[0110] Where f' represents the data-driven estimation algorithm used to establish the ΔSOH data-driven estimation model, Ψ ΔSOH Represents the constructed features for ΔSOH data-driven modeling.

[0111] Based on formula (2), the differential value ΔSOH of SOH is used as the modeling target. By modeling ΔSOH with a shorter modeling time scale and a smaller range as the target, it helps to reduce the probability of large estimation deviations in the traditional SOH data-driven estimation method and increase the stability of the SOH estimation results. In addition, batteries age as they are used, so the physical meaning of the ΔSOH data-driven estimation model is clear. However, although the ΔSOH data-driven estimation method will not have large estimation deviations at certain moments, there is also an inevitable problem of error accumulation. Therefore, it is necessary to update the SOH value obtained by the ΔSOH data-driven estimation model at a specific point.

[0112] Based on the above analysis, a hybrid data-driven estimation model based on SOH and ΔSOH is designed. That is, a SOH data-driven estimation model and a ΔSOH data-driven estimation model are established on both the vehicle side and the cloud side. The characteristics are as follows:

[0113]

[0114] The SOH data-driven estimation model is updated by monitoring the full-charging stage, where the full-charging stage refers to an interval state of charge ΔSOC ≥ 80% during the charging process. The SOH estimation result obtained by the updated SOH data-driven estimation model is used to update the SOH estimation value obtained by the ΔSOH data-driven estimation model, and the SOH estimation value obtained by the ΔSOH data-driven estimation model is used as the SOH estimation result in the non-full-charging stage.

[0115] Step 2: Construction of aging characteristics based on statistical characteristic indicators

[0116] In the real world, the charging method of electric vehicles may involve complex multi-stage constant current charging. The variables such as battery, voltage, and current are affected by measurement noise and communication quality, resulting in partial data loss. This may make the traditional aging feature construction method based on the charging voltage curve no longer applicable. Therefore, based on the actual vehicle operating condition data, the statistical characteristic indicators of the parameters constructed based on voltage and current are used to construct aging features for SOH data-driven modeling and ΔSOH data-driven modeling, taking into account the data storage and computing capabilities of the vehicle and the cloud.

[0117] S2.1. Aging feature construction for SOH data-driven modeling

[0118] With the continuous use and aging of the battery, the active lithium ions in the battery are continuously lost, the active materials are continuously decomposed, the SEI film is continuously grown, etc., and the internal resistance of the battery continues to increase. This makes the battery response parameters such as voltage and temperature become more unstable under big data, resulting in significant changes in the discreteness, confusion and distribution characteristics of these measurable response parameters as the battery ages.

[0119] a) Selection of statistical characteristic indicators for constructing aging characteristics

[0120] In the SOH data-driven modeling, on the basis of selecting the standard deviation and coefficient of variation representing the degree of data dispersion, and the skewness, kurtosis and peak-to-skew ratio representing the data distribution characteristics, the mean and quartile representing the degree of data dispersion and the discrete entropy representing the degree of data chaos are also selected, a total of eight basic statistical characteristic indicators are used to construct the aging characteristics. In addition, in order to increase the robustness of the SOH data-driven modeling, on the basis of the above eight basic statistical characteristic indicators, different combinations of statistical characteristic indicators can be added to provide more diverse aging characteristics for the SOH data-driven modeling, such as the combined statistical characteristic indicators constructed by standard deviation*kurtosis, standard deviation / kurtosis, standard deviation*ln(kurtosis), standard deviation / ln(kurtosis), ln(standard deviation)*kurtosis, ln(standard deviation) / kurtosis, standard deviation*peak-to-skew ratio, coefficient of variation*ln(peak-to-skew ratio) and coefficient of variation / ln(peak-to-skew ratio);

[0121] b) Parameter selection for constructing aging features

[0122] As the battery ages, the internal resistance of the battery increases, resulting in increasing fluctuations and confusion in the battery response voltage under actual vehicle operating conditions. Based on the combined parameters of voltage divided by current, the specific aging characteristic value is calculated according to the aging characteristic constructed in S2.1 a). Considering that the driving current value under actual vehicle operating conditions may be 0, this will cause the combined signal value of voltage divided by current to be invalid. Therefore, in order to ensure the validity of the calculated aging characteristic value, the current is input into the corresponding exponential function equation. The parameters used to construct the aging characteristic are as follows:

[0123]

[0124] Where U t Represents the battery terminal voltage, I t Indicates the battery current;

[0125] c) Specific calculation method of aging characteristics

[0126] The more actual vehicle operating condition data used to calculate the aging characteristics, the less the construction of the aging characteristics will be affected by measurement noise, missing values, specific operating conditions, etc., the more aging information it contains, and the more stable the construction of the aging characteristics. Therefore, according to the statistical characteristic indicators for constructing the aging characteristics selected in S2.1 a) and the parameters for constructing the aging characteristics selected in S2.1 b), the aging characteristics are calculated based on the accumulation of actual vehicle operating condition data as follows:

[0127] l SOH,k =F ea,j (D a ),D a ={X i ,i=1,...,k} (5)

[0128] In the formula, l SOH,k represents the aging characteristic value calculated at the kth moment for SOH estimation, F ea,j represents the selected statistical feature index, j = 1, ..., M, M is the number of statistical feature indicators, D a Represents the actual vehicle operating condition data accumulated from the first moment to the current k moment;

[0129] d) Online incremental calculation of aging characteristics

[0130] Although the larger the amount of accumulated data used to construct the aging features, the more stable and accurate the constructed aging features are. However, considering the limited data storage and computing capabilities on the vehicle side, the constructed aging features need to have real-time online incremental computing capabilities, that is, the expected effective aging features can be calculated without storing data. Since quartiles and discrete entropy must be based on accumulated data to calculate aging feature values. Therefore, for SOH data-driven modeling, the cloud constructs aging features based on eight basic statistical feature indicators and combined statistical feature indicators; due to the limited data storage and computing capabilities on the vehicle side, the statistical feature indicators used for aging feature construction on the vehicle side need to have effective online incremental calculation functions. Therefore, the vehicle side constructs aging features based on the mean, standard deviation, coefficient of variation, skewness, kurtosis, and peak-to-skewness ratio with online incremental calculation functions. The online incremental calculation equations of the six related aging features used for vehicle-side SOH data-driven modeling are as follows:

[0131] ①Online incremental calculation equation of aging characteristics constructed based on the mean value:

[0132]

[0133] in,

[0134]

[0135] In the formula, Characterizes the mean index calculated at the nth moment, Represents the mean index calculated at the n+1th moment, X n+1 is the actual vehicle operating condition data at the n+1th moment, and n is the number of data used for feature calculation from the 1st moment to the current moment;

[0136] ②Online incremental calculation equation of aging characteristics constructed based on standard deviation:

[0137]

[0138] in,

[0139]

[0140] In the formula, Characterizes the standard deviation index calculated at the nth moment, Characterizes the standard deviation index calculated at the n+1th moment;

[0141] ③Online incremental calculation equation of aging characteristics constructed based on difference coefficient:

[0142]

[0143] in,

[0144]

[0145] In the formula, Characterizes the difference coefficient index calculated at the nth moment, Characterizes the difference coefficient index calculated at the n+1th moment;

[0146] ④Online incremental calculation equation based on aging characteristics constructed by skewness:

[0147]

[0148] in,

[0149]

[0150] In the formula, Characterizes the skewness index calculated at the nth moment, Characterizes the skewness index calculated at the n+1th moment;

[0151] ⑤Online incremental calculation equation based on aging characteristics constructed by kurtosis:

[0152]

[0153] in,

[0154]

[0155] In the formula, Characterizes the kurtosis index calculated at the nth moment, Characterizes the kurtosis index calculated at the n+1th moment;

[0156] ⑥Online incremental calculation equation based on aging characteristics constructed by peak deviation ratio:

[0157]

[0158] In the formula, Characterizes the peak deviation ratio index calculated at the n+1th moment.

[0159] S2.2. Construction of aging features for ΔSOH data-driven modeling

[0160] ΔSOH characterizes battery loss, which is closely related to factors such as the current size when the battery is in use, the frequency of current charging and discharging fluctuations, and temperature. As the only source of battery energy consumption, electric vehicles are closely related to vehicle parameters such as the current size when the battery is in use, the frequency of current charging and discharging fluctuations, and vehicle speed, changes in the longitude and latitude of the vehicle's position, and the travel value of the acceleration or brake pedal. In addition, the battery usage time (measured from the battery's perspective) and the vehicle's mileage (measured from the vehicle's perspective) also directly determine the size of ΔSOH.

[0161] a) Selection of statistical characteristic indicators for constructing aging characteristics

[0162] Since the current size and the fluctuation frequency of current charging and discharging during battery use determine the loss of battery capacity, the eight basic statistical characteristic indicators and combined statistical characteristic indicators selected in S2.1 can still be used to construct aging characteristics for ΔSOH data-driven modeling;

[0163] b) Parameter selection for constructing aging features

[0164] From the perspective of the battery, the current size when the battery is in use, and the frequency of current charging and discharging fluctuations are reflected in the changes in the battery's response voltage and temperature parameters. From the perspective of the vehicle, the current size when the battery is in use, and the frequency of current charging and discharging fluctuations are determined by the actual vehicle operating conditions. Vehicle parameters such as vehicle speed, changes in longitude and latitude of the vehicle's location, and acceleration or brake pedal travel values ​​can all be used to measure battery aging. Based on the characteristics of the vehicle and cloud sides, the vehicle-side ΔSOH data-driven modeling selects measurable parameters of the battery including voltage, current, and temperature to construct aging characteristics, and the cloud-side ΔSOH data-driven modeling selects measurable parameters of the battery including voltage, current, and temperature, and measurable parameters of the vehicle including vehicle speed, longitude of the vehicle's location, latitude of the vehicle's location, accelerator pedal travel value, and brake pedal travel value to construct aging characteristics;

[0165] c) Specific calculation method of aging characteristics

[0166] ΔSOH can characterize the loss of battery capacity over a period of time. Therefore, the specific calculation method of aging characteristics used for ΔSOH data-driven modeling no longer requires accumulated data calculation like SOH data-driven modeling. Instead, for ΔSOH data-driven modeling, the aging characteristics are calculated based on the actual vehicle operating condition data with a fixed mileage as follows:

[0167] l ΔSOH,k =F ea,j (D' a ),D' a ={X i ,i=km,...,k} (17)

[0168] In the formula, l ΔSOH,k represents the aging characteristic value calculated at the kth moment for ΔSOH estimation, D' a It represents the actual vehicle operating condition data accumulated from time km to the current time k based on a fixed mileage, and m represents the length of the data window.

[0169] In actual vehicle operation conditions, factors such as charging current and charging time will affect the battery reference capacity calculated in the full charge stage, resulting in large fluctuations in battery capacity on a shorter time scale. If the calculated reference capacity with obvious fluctuation characteristics is directly obtained by differential without any processing to obtain the ΔSOH target value, the capacity calculation error caused by factors such as reference capacity and capacity regeneration will affect the ΔSOH calculation value for long-term battery aging, which makes the calculated ΔSOH value possibly positive or negative, which is inconsistent with the actual physical meaning of ΔSOH. Therefore, for ΔSOH data-driven modeling, the vehicle's mileage and the SOH reference value obtained based on the battery reference capacity can be fitted based on the polynomial fitting equation, where the battery reference capacity is calculated based on the full charge stage of the battery as follows:

[0170]

[0171] In the formula, Cap ref It represents the reference capacity of the battery, ΔQ represents the charging capacity of the battery during the charging process, and ΔSOC represents the interval state of charge during the charging process. When calculating the reference capacity of the battery, it is necessary to ensure that the ΔSOC of the battery charging process is greater than or equal to 80%.

[0172] Therefore, on the basis of obtaining the smooth attenuation curve of battery aging, the ΔSOH target modeling value with real physical meaning is obtained as follows:

[0173] ΔSOH=diff(Poly(Μ mileage ,SOH ref )) (19)

[0174] In the formula, diff represents differential calculation, Poly represents the polynomial fitting equation, and M mileage Indicates the vehicle mileage value, SOH ref It represents the SOH reference value obtained based on the battery reference capacity, which is calculated as follows:

[0175]

[0176] In the formula, Cap nor Indicates the nominal capacity of the battery when it is not aged;

[0177] d) Online incremental calculation of aging characteristics

[0178] Since the aging features constructed by ΔSOH data-driven modeling are the same as those constructed by SOH data-driven modeling, the online incremental calculation equations of the relevant aging features in formulas (6) to (16) are still applicable to the online incremental calculation of the aging features of ΔSOH data-driven modeling.

[0179] Step 3: Vehicle-Cloud Collaborative SOH Estimation

[0180] Combining the strategic principles of vehicle-cloud collaborative SOH estimation Figure 2 As shown, the details are as follows:

[0181] S3.1. Self-monitoring and model updating in the car-cloud

[0182] Considering the applicability of the SOH data-driven estimation model established on heterogeneous data, the vehicle side and the cloud side need to design self-monitoring and model update strategies to ensure the validity of the SOH estimation results obtained by the vehicle side and the cloud side. Since a more accurate reference capacity can be obtained during the full charging stage, a more accurate SOH reference value can be obtained. Therefore, the SOH reference value SOH can be obtained based on the obtained SOH reference value SOH. ref Monitor the effectiveness of the SOH data-driven estimation model and the ΔSOH data-driven estimation model established on the vehicle side and the cloud side. If the following conditions are met, the SOH data-driven estimation model and the ΔSOH data-driven estimation model need to be updated.

[0183] |SOH ref -SOH Cloud / SOH Vehicle |>δ (21)

[0184] In the formula, SOH Cloud Represents the estimated SOH value obtained by the cloud. Vehicle It represents the estimated SOH value obtained by the vehicle end, and δ is the set threshold value of the abnormal SOH estimation result, which can be taken as 2% or 3%.

[0185] Since the cloud has strong data storage and computing capabilities, the SOH and ΔSOH data-driven estimation models established on the cloud are updated based on batch learning. Since the data storage and computing capabilities of the vehicle are limited, the vehicle can perform online incremental learning based on the OSELM algorithm.

[0186] In addition, the vehicle and cloud also need to monitor the SOH estimation results obtained based on the established SOH data-driven estimation model. Compared with the SOH estimation results obtained based on the established ΔSOH data-driven estimation model If the following conditions are met, it means that the SOH estimation results obtained are significantly different between the vehicle side and the cloud side.

[0187]

[0188] In this case, it is necessary to determine the time interval between the current moment and the last model update. When the time interval is less than or equal to half a month, the SOH estimation result based on ΔSOH data-driven modeling can be used as the SOH estimation result at the current moment, because the SOH estimation result based on ΔSOH data-driven modeling is more reliable in a shorter time interval after the model update. Conversely, if the time interval is greater than half a month, the average of the SOH estimation result obtained based on SOH data-driven modeling and the SOH estimation result obtained based on ΔSOH data-driven modeling can be used as the SOH estimation result at that moment.

[0189] S3.2. Vehicle-cloud mutual monitoring and collaborative SOH estimation strategy

[0190] In addition to monitoring and model updating on the vehicle side and the cloud side, a mutual monitoring strategy between the vehicle side and the cloud side needs to be designed to further ensure the validity of the SOH estimation results obtained by the vehicle side and the cloud side. The design of a collaborative SOH estimation strategy is reflected in two aspects: mutual monitoring of the SOH estimation results between the vehicle side and the cloud side, and mutual feedback and update of the measurable parameters voltage, current and temperature data of the battery between the vehicle side and the cloud side.

[0191] First, when it is impossible to experience the full charge stage of the battery (ΔSOC ≥ 80%) to obtain an accurate SOH reference value SOH ref When the vehicle-side and cloud-side models are fully charged, the vehicle-side and cloud-side models need to undergo a full charge phase for self-monitoring and model update. The difference between the SOH estimation results obtained by the vehicle-side and the SOH estimation results obtained by the cloud-side is compared with the set threshold, as shown in the following formula:

[0192] |SOH Cloud -SOH Vehicle |>δ (23)

[0193] If formula (23) is satisfied, it is necessary to determine whether the model established on the cloud side has been effectively updated during the full charging stage. If it has been updated, the SOH data-driven estimation model on the vehicle side is updated according to the SOH estimation result obtained on the cloud side. Otherwise, the average value of the SOH estimation results obtained on the vehicle side and the cloud side is taken as the SOH estimation value at the current moment. If formula (23) is not satisfied, it means that the model states used to estimate SOH on the vehicle side and the cloud side are basically the same, and there is no need to start the model update on the vehicle side and the cloud side.

[0194] In addition, the vehicle side can collect more refined battery response data to achieve more accurate SOC estimation. These accurate SOC estimation results can provide accurate calculation information for the calculation of cloud reference capacity, thereby obtaining an accurate SOH reference value.

[0195] Example

[0196] This embodiment is verified based on the actual vehicle operating condition data of a real-world pure electric vehicle (vehicle 1). The battery type of this pure electric vehicle is a ternary battery with a nominal capacity of 150Ah. The data collection time is 10s, and the collected parameters include: voltage, current, temperature, speed, longitude, latitude, acceleration pedal travel value and brake pedal travel value. The time of data collection for this vehicle is: 2022 / 1 / 7 to 2022 / 12 / 31, and the mileage is 161412.2m.

[0197] The aging characteristics constructed based on the selected eight basic statistical characteristic indicators and the combined statistical characteristic indicators are compared with the obtained SOH reference value (SOH ref ) are shown in Table 1 below:

[0198] Table 1 Correlation between aging characteristics and SOH reference values

[0199]

[0200] Figure 3 and Figure 4 The SOH estimation results obtained based on SOH data-driven modeling and ΔSOH data-driven modeling under the aging characteristics constructed by eight basic statistical characteristic indicators and the combined statistical characteristic indicators are given respectively.

[0201] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other forms of assembly without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations within the meaning and range of equivalents of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0202] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A vehicle-cloud collaborative SOH estimation method for real vehicle operation data, characterized by: The following steps are involved: Step 1: Hybrid data-driven modeling of SOH and ΔSOH The SOH data-driven estimation model is characterized as follows: Μ SOH =f(Ψ SOH ,SOH) (1) Where f represents the data-driven estimation algorithm used to establish the SOH data-driven estimation model, Ψ SOH Represents the constructed modeling features for SOH data-driven; Taking the difference value ΔSOH of SOH as the modeling target, the ΔSOH data-driven estimation model is characterized as follows: M ΔSOH =f'(Ψ ΔSOH ,ΔSOH) (2) Where f' represents the data-driven estimation algorithm used to establish the ΔSOH data-driven estimation model, Ψ ΔSOH represents the constructed modeling features for ΔSOH data-driven; A hybrid data-driven estimation model based on SOH and ΔSOH is designed. The SOH data-driven estimation model and the ΔSOH data-driven estimation model are established on both the vehicle side and the cloud side. The characteristics are as follows: The SOH data-driven estimation model is updated by monitoring the full-charge stage, wherein the full-charge stage refers to the interval state of charge ΔSOC ≥ 80% during the charging process, and the SOH estimation result obtained by the updated SOH data-driven estimation model is used to update the SOH estimation value obtained by the ΔSOH data-driven estimation model, and the SOH estimation value obtained by the ΔSOH data-driven estimation model is used as the SOH estimation result in the non-full-charge stage; Step 2: Construction of aging characteristics based on statistical characteristic indicators S2.

1. Aging feature construction for SOH data-driven modeling a) Selection of statistical characteristic indicators for constructing aging characteristics In the SOH data-driven modeling, in addition to selecting the standard deviation and coefficient of variation representing the degree of data dispersion, and the skewness, kurtosis, and peak-to-skewness ratio representing the data distribution characteristics, the mean and quartile representing the degree of data dispersion, and the discrete entropy representing the degree of data chaos are also selected. A total of eight basic statistical characteristic indicators are selected to construct the aging characteristics; b) Parameter selection for constructing aging features To ensure the validity of the calculated aging characteristic value, the current is input into the exponential function equation. The parameters used to construct the aging characteristic are as follows: Where U t Represents the battery terminal voltage, I t Indicates the battery current; c) Specific calculation method of aging characteristics According to the statistical characteristic index for constructing the aging characteristic selected in S2.1 a) and the parameters for constructing the aging characteristic selected in S2.1 b), the aging characteristic is calculated based on the accumulation of the actual vehicle operating condition data as follows: l SOH,k =F ea,j (D a ),D a ={X i ,i=1,...,k} (5) In the formula, l SOH,k represents the aging characteristic value calculated at the kth moment for SOH estimation, F ea,j represents the selected statistical feature index, j = 1, ..., M, M is the number of statistical feature indicators, D a Represents the actual vehicle operating condition data accumulated from the first moment to the current k moment; d) Online incremental calculation of aging characteristics The vehicle side constructs aging features based on the mean, standard deviation, coefficient of variation, skewness, kurtosis, and peak-to-skewness ratio with online incremental calculation function. The online incremental calculation equations of the six relevant aging features used for vehicle side SOH data-driven modeling are as follows: ①Online incremental calculation equation of aging characteristics constructed based on the mean value: in, In the formula, Characterizes the mean index calculated at the nth moment, Represents the mean index calculated at the n+1th moment, X n+1 is the actual vehicle operating condition data at the n+1th moment, and n is the number of data used for feature calculation from the 1st moment to the current moment; ②Online incremental calculation equation of aging characteristics constructed based on standard deviation: in, In the formula, Characterizes the standard deviation index calculated at the nth moment, Characterizes the standard deviation index calculated at the n+1th moment; ③Online incremental calculation equation of aging characteristics constructed based on difference coefficient: in, In the formula, Characterizes the difference coefficient index calculated at the nth moment, Characterizes the difference coefficient index calculated at the n+1th moment; ④Online incremental calculation equation based on aging characteristics constructed by skewness: in, In the formula, Characterizes the skewness index calculated at the nth moment, Characterizes the skewness index calculated at the n+1th moment; ⑤Online incremental calculation equation based on aging characteristics constructed by kurtosis: in, In the formula, Characterizes the kurtosis index calculated at the nth moment, Characterizes the kurtosis index calculated at the n+1th moment; ⑥Online incremental calculation equation based on aging characteristics constructed by peak deviation ratio: In the formula, Characterizes the peak deviation ratio index calculated at the n+1th moment; S2.

2. Construction of aging features for ΔSOH data-driven modeling a) Selection of statistical characteristic indicators for constructing aging characteristics Use the statistical characteristic indicators selected in S2.1 a) to construct aging characteristics for ΔSOH data-driven modeling; b) Parameter selection for constructing aging features The vehicle-side ΔSOH data-driven modeling selects the measurable parameters of the battery including voltage, current and temperature to construct the aging characteristics. The cloud-side ΔSOH data-driven modeling selects the measurable parameters of the battery including voltage, current and temperature and the measurable parameters of the vehicle including vehicle speed, longitude of vehicle location, latitude of vehicle location, accelerator pedal travel value and brake pedal travel value to construct the aging characteristics. c) Specific calculation method of aging characteristics For ΔSOH data-driven modeling, the aging characteristics are calculated based on the actual vehicle operating condition data with fixed mileage as follows: l ΔSOH,k =F ea,j (D a '),D a '={X i ,i=k-m,...,k} (17) In the formula, l ΔSOH,k represents the aging characteristic value calculated at the kth moment for ΔSOH estimation, D' a It represents the actual vehicle operating condition data accumulated from the fixed mileage, i.e., from the km moment to the current k moment, and m represents the data window length; For ΔSOH data-driven modeling, the vehicle mileage and the SOH reference value obtained based on the battery reference capacity are fitted based on the polynomial fitting equation, where the battery reference capacity is calculated based on the full charge stage of the battery as follows: In the formula, Cap ref represents the reference capacity of the battery, and ΔQ represents the charging capacity of the battery during the charging process; Therefore, on the basis of obtaining the smooth attenuation curve of battery aging, the ΔSOH target modeling value with real physical meaning is obtained as follows: ΔSOH=diff(Poly(Μ mileage ,SOH ref )) (19) In the formula, diff represents differential calculation, Poly represents the polynomial fitting equation, and M mileage Indicates the vehicle mileage value, SOH ref It represents the SOH reference value obtained based on the battery reference capacity, which is calculated as follows: In the formula, Cap nor Indicates the nominal capacity of the battery when it is not aged; d) Online incremental calculation of aging characteristics Since the aging features constructed by ΔSOH data-driven modeling are the same as those constructed by SOH data-driven modeling, the online incremental calculation equations of the relevant aging features in formulas (6) to (16) are also used for the online incremental calculation of the aging features of ΔSOH data-driven modeling; Step 3: Vehicle-Cloud Collaborative SOH Estimation S3.

1. Self-monitoring and model updating in the car-cloud The effectiveness of the SOH data-driven estimation model and the ΔSOH data-driven estimation model established on the vehicle side and the cloud side is monitored according to the obtained SOH reference value. If the following conditions are met, the SOH data-driven estimation model and the ΔSOH data-driven estimation model need to be updated; SOH ref -SOH Cloud / SOH Vehicle |>δ (21) In the formula, SOH Cloud Represents the estimated SOH value obtained by the cloud. Vehicle represents the SOH estimation value obtained by the vehicle end, and δ is the set threshold for abnormal SOH estimation result; The vehicle and the cloud also need to monitor the SOH estimation results obtained based on the established SOH data-driven estimation model. MSOH Compared with the SOH estimation results obtained based on the established ΔSOH data-driven estimation model, SOH MΔSOH If the following conditions are met, it means that the SOH estimation results obtained are significantly different between the vehicle side and the cloud side; SOH MSOH -SOH MΔSOH |>δ (22) In this case, it is necessary to determine the time interval between the current moment and the last model update. When the time interval is less than or equal to half a month, the SOH estimation result based on ΔSOH data-driven modeling can be used as the SOH estimation result at the current moment. On the contrary, if the time interval is greater than half a month, the average of the SOH estimation result based on SOH data-driven modeling and the SOH estimation result based on ΔSOH data-driven modeling is used as the SOH estimation result at that moment. S3.

2. Vehicle-cloud mutual monitoring and collaborative SOH estimation strategy In addition to monitoring and model updating on the vehicle side and the cloud side, a mutual monitoring strategy between the vehicle side and the cloud side needs to be designed to further ensure the validity of the SOH estimation results obtained by the vehicle side and the cloud side. First, when it is impossible to obtain an accurate SOH reference value by fully charging the battery, a full charging phase is required to self-monitor and update the models on the vehicle side and the cloud side, and compare the difference between the SOH estimation results obtained on the vehicle side and the SOH estimation results obtained on the cloud side with the set threshold, as shown in the following formula: SOH Cloud -SOH Vehicle |>δ (23) If formula (23) is satisfied, it is necessary to determine whether the model established on the cloud side has been updated during the full charging stage. If it has been updated, the SOH data-driven estimation model on the vehicle side is updated according to the SOH estimation result obtained on the cloud side. Otherwise, the average value of the SOH estimation results obtained on the vehicle side and the cloud side is taken as the SOH estimation value at the current moment. If formula (23) is not satisfied, there is no need to start the model update on the vehicle side and the cloud side.

2. The vehicle-cloud collaborative SOH estimation method for real vehicle operation data according to claim 1 is characterized by: When selecting statistical characteristic indicators for constructing aging characteristics in a) of S2.1, on the basis of the eight basic statistical characteristic indicators, combined statistical characteristic indicators are added, including: standard deviation*kurtosis, standard deviation / kurtosis, standard deviation*ln(kurtosis), standard deviation / ln(kurtosis), ln(standard deviation)*kurtosis, ln(standard deviation) / kurtosis, standard deviation*peak-to-peak ratio, coefficient of difference*ln(peak-to-peak ratio) and coefficient of difference / ln(peak-to-peak ratio).

3. The vehicle-cloud collaborative SOH estimation method for real vehicle operation data according to claim 1 is characterized by: The set threshold δ for abnormal SOH estimation result in S3.1 is set to 2% or 3%.