CN114740356A - Power battery residual life indirect prediction method of similar voltage sequence - Google Patents

Power battery residual life indirect prediction method of similar voltage sequence Download PDF

Info

Publication number
CN114740356A
CN114740356A CN202110020952.XA CN202110020952A CN114740356A CN 114740356 A CN114740356 A CN 114740356A CN 202110020952 A CN202110020952 A CN 202110020952A CN 114740356 A CN114740356 A CN 114740356A
Authority
CN
China
Prior art keywords
voltage
discharge
battery
sequence
power battery
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110020952.XA
Other languages
Chinese (zh)
Inventor
叶敏
魏孟
王桥
武晨光
麻玉川
张佳乐
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changan University
Original Assignee
Changan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Changan University filed Critical Changan University
Priority to CN202110020952.XA priority Critical patent/CN114740356A/en
Publication of CN114740356A publication Critical patent/CN114740356A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/70—Energy storage systems for electromobility, e.g. batteries

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Secondary Cells (AREA)
  • Tests Of Electric Status Of Batteries (AREA)

Abstract

The invention discloses a method for indirectly predicting the residual life of a power battery with similar voltage sequences. Real-time and accurate RUL prediction is a precondition for ensuring safe and effective work of the electric vehicle; however, the direct health factor of the lithium ion battery is difficult to realize on-line measurement, and aiming at the phenomenon, firstly, the discharge voltage cut-off time is provided as a new indirect establishment factor by analyzing the discharge cycle data of the battery and introducing a grey correlation theory; secondly, aiming at the problems that the traditional single historical cycle data has accumulated errors and can not accurately reflect the unconventional degradation of the battery, similarity analysis is carried out through a twin neural network to obtain a similarity discharge cut-off voltage sequence; and finally, introducing a similarity discharge cut-off voltage sequence on the basis of the monomer discharge cut-off voltage sequence, establishing a Gaussian mixture model to obtain real-time and accurate RUL and giving a predicted confidence interval.

Description

Power battery residual life indirect prediction method of similar voltage sequence
Technical Field
The invention belongs to the technical field of battery fault prediction and health diagnosis, and relates to a method for indirectly predicting the residual life of a power battery with similar voltage sequences.
Background
With the increasing prominence of environmental pollution and energy shortage problems, electric vehicles are more and more closely concerned by relevant departments and enterprise researchers. Lithium ion batteries are widely used in the field of electric vehicles due to their advantages of high energy density, long cycle life, low cost, etc. Accurate prediction of the Remaining battery life (RUL) is a basic premise for safe operation of the electric vehicle and is also an important support factor of the battery management system. Fault Prediction and Health Management (PHM) of a battery system is one of the key technologies of an electric vehicle, wherein real-time and accurate RUL prediction is a precondition for ensuring stability, effectiveness and safety of a power battery. However, since direct health factors such as capacity and the like are difficult to realize on-line real-time measurement and degradation factors are mutually coupled, it is of great significance to provide a simple and effective method for predicting the remaining life of the lithium ion battery. The invention provides a method for indirectly predicting the residual life of a power battery with similar voltage sequences. Currently, the prediction of the power battery RUL is generally divided into two categories, including mechanism modeling and data-driven. The mechanism model method identifies the corresponding relationship between the measurable value and the health index by establishing a physical model of the lithium ion battery degradation. Common methods include an exponential model method, a linear model method, a logarithmic model method, a Kalman filtering method and the like. The data-driven method is characterized in that degradation information of the battery is mined and a degradation model of the battery is established by analyzing battery state monitoring data. Common methods include static neural networks, dynamic neural networks, support vector machines, gaussian process regression, and the like. Under the actual complex vehicle-mounted working condition: the capacity decline of the battery shows an unconventional change trend, and the real-time measurement difficulty of the direct health factor is large. Therefore, according to the battery discharge cycle data, a grey correlation theory is introduced, and the discharge voltage cut-off time is proposed as a new indirect health factor. Aiming at the problem that the error accumulation generated by the traditional single historical cycle data training cannot accurately reflect the unconventional degradation trend of the battery, the invention introduces a voltage similarity sequence and adopts a twin neural network to learn the integral degradation characteristics of a multi-sequence battery to describe the mutation information in the battery degradation process. On the basis of the monomer discharge cut-off voltage sequence and the similar discharge cut-off voltage sequence, a Gaussian mixture regression model is adopted to establish a prediction model of the residual life of the power battery and obtain a confidence interval.
Disclosure of Invention
The invention provides a power battery residual life indirect prediction method with similar voltage sequences in order to obtain online measurement health factors and obtain a reliable confidence interval.
Because the vehicle-mounted working condition is complicated and changeable, the actual running state, the operation condition, the external interference and the like of the vehicle lead the actual capacity of the power battery to have unstable variation trend. Meanwhile, capacity and internal resistance as direct health factors are difficult to realize on-line measurement. Therefore, it is very important to construct indirect health factors that can be actually measured.
By analyzing the discharging condition of the battery in the actual vehicle-mounted process, the average discharging current time, the discharging voltage cut-off time and the discharging temperature peak value time are provided as indirect health factors. The method comprises the following steps of considering the actual vehicle-mounted working condition, and simultaneously obtaining the following results through grey correlation analysis: the voltage parameter is used as a direct reference basis of a user, and the practicability is high. Therefore, the discharge voltage off-time is adopted as a new health factor.
In a traditional neural network model, the power battery state parameters of historical cycles are generally adopted to predict the result. When the state parameters are predicted, prediction error accumulation is generated, and the unconventional decline of the battery in the actual vehicle-mounted process cannot be predicted, so that the RUL prediction effect is influenced. In order to solve the irregular degradation phenomenon of the battery in the degradation process, the method comprises the following steps: the capacity regeneration phenomenon is characterized in that a twin neural network-based similarity theory is adopted, and a multi-sequence state is introduced to serve as training data of a model.
By analyzing the Battery Data Set provided by the national aeronautics and astronautics NASA PCoE research center, as shown in fig. 1, although specific differences occur in the degradation process of 4 18650 lithium ion batteries (B5, B6, B7, B18), the degradation process is similar between each Battery cell, in order to solve unconventional degradation trends, such as: the capacity regeneration phenomenon and the error accumulation phenomenon of single-sequence prediction reflect the unconventional phenomenon in the battery degradation process by learning the overall degradation characteristics of multiple sequences so as to obtain more accurate prediction of the remaining life of the battery.
As can be seen from fig. 1, the overall degradation characteristics and the local capacity regeneration phenomena of the four batteries have high similarities. Meanwhile, the similarity of the discharge voltage cut-off sequence is analyzed by adopting a grey correlation theory. And finally, carrying out similarity quantitative analysis on the voltage sequences of the 4 batteries based on the twin neural network.
The structure of the twin neural network is shown in fig. 2. The twin neural network adopts a Contrast Loss Function (CLF), the contrast Loss Function can efficiently process the direct relation of input data, and the CLF expression is as follows:
Figure BSA0000229611850000031
Figure BSA0000229611850000032
wherein DWRepresenting input samples XiAnd XjP represents the characteristic dimension of the input sample, and Y represents the label of whether the sample matches.
If Y ═ 1 indicates that the samples are similar, then the loss function is:
Figure BSA0000229611850000033
if sample Y is 0, the loss function is:
Figure BSA0000229611850000034
by analyzing the discharge historical cycle data of the battery and considering the practicability of the actual vehicle-mounted working condition, an indirect health factor based on the discharge cut-off time is established, a twin neural network is adopted for similarity analysis, a voltage similarity theory is introduced, and a monomer discharge voltage cut-off time sequence and a voltage similarity sequence are adopted to represent the overall degradation characteristic of the battery and the unconventional degradation phenomenon. Under the actual vehicle-mounted complex working condition, the confidence coefficient of obtaining the RUL of the lithium ion battery is more practical than the traditional RUL point estimation. Therefore, the method establishes the RUL of the Gaussian mixture prediction lithium ion battery and obtains a confidence interval on the basis of the monomer discharge cut-off voltage time sequence and the similar voltage sequence.
On the basis of the monomer discharge cut-off voltage time sequence and the similar voltage sequence, taking the first 50 percent of the degradation period as training data and the last 50 percent as test data; and (5) carrying out cluster analysis on the battery data by adopting a K-means clustering algorithm. The data for the lithium ion battery includes output and input data cell discharge cutoff voltage sequences, similar discharge cutoff voltage sequences, and battery RUL. Establishing a joint density function of a Gaussian mixture model:
Figure BSA0000229611850000041
solving the hyperparameter of Gaussian mixture by EM algorithm to obtain the optimal hyperparameter
Q＝{{α1，μ1，∑1}，{α2，μ2，∑2}，...，{αn，μn，∑n}}
Calculating the posterior probability of each Gaussian component test set by outputting test set data:
μj(x)＝μjy+∑jyx∑jxx -1(x-μjx)
∑j(x)＝∑jyy-∑jyx∑jxx -1∑jxy
predicting RUL and confidence intervals by gaussian mixed regression prediction output
Figure BSA0000229611850000042
Figure BSA0000229611850000043
The invention provides a power battery residual life indirect prediction method with similar voltage sequences by analyzing state parameters among battery packs and aiming at the problems that direct health factors are difficult to measure on line under the actual vehicle-mounted working condition and the traditional neural network is used for point estimation for prediction.
Drawings
Fig. 1 is a graph of capacity fade of a battery pack according to the present invention;
FIG. 2 is a structural framework diagram of a twin neural network according to the present invention;
FIG. 3 is a structural diagram of a method for predicting remaining battery life according to the present invention;
Detailed Description
Step 1: the experimental acquisition of the state parameters of the vehicle-mounted power battery comprises the following steps: current, voltage, temperature, and capacity;
and 2, step: by analyzing discharge cycle data of the power battery, indirect health factors are extracted: discharge cutoff time, average discharge current time, and discharge temperature peak time;
and step 3: analyzing the correlation between the indirect factor and the battery degradation characteristic through a grey correlation theory, considering the actual vehicle-mounted working condition, and selecting a discharge cut-off voltage as a new health factor;
and 4, step 4: in order to obtain the confidence of the RUL prediction, a Gaussian mixture model is adopted to establish a prediction model of the RUL of the lithium ion battery.
And 5: and (3) taking the inevitable error accumulation of the conventional single-sequence voltage historical data cycle and the irregular change in the degradation process into consideration, introducing a voltage similarity sequence as an auxiliary training.
Step 6: and (3) carrying out similarity analysis on the direct discharge cut-off voltage time of the battery pack by adopting a twin neural network to obtain a monomer discharge cut-off voltage sequence and a similar discharge voltage sequence.
And 7: and substituting the obtained monomer discharge cut-off voltage sequence and the similar discharge voltage sequence into a Gaussian mixture model to obtain RUL prediction and obtain a confidence interval.
Firstly, battery discharge data is analyzed, discharge voltage cut-off time is provided as a new health factor in consideration of actual vehicle-mounted working conditions, the direct correlation between the lithium ion battery capacity and the discharge cut-off voltage is analyzed through a grey correlation theory, a voltage similarity theory is introduced, the voltage characteristics of the lithium ion battery are analyzed through a twin neural network, principle support is provided for the voltage similarity theory, and multi-sequence degradation characteristics are introduced. On the basis, in order to obtain the service life prediction confidence of the RUL of the lithium ion battery, a Gaussian mixture model is established by considering the single discharge cut-off voltage sequence and the similar discharge cut-off voltage sequence to predict the RUL of the lithium ion battery and give a confidence interval.

Claims (4)

1. A method for indirectly predicting the residual life of a power battery with similar voltage sequences is characterized by comprising the following steps:
according to the discharge data characteristics of the power battery, under the condition of considering the actual vehicle-mounted working condition, the discharge voltage cut-off time is provided as an indirect health factor for predicting the residual life of the power battery;
considering the problem that errors generated when historical discharge cutoff voltage cycle data are accumulated and abnormal degradation of the battery cannot be accurately reflected, the auxiliary similar voltage sequence and the single historical cycle voltage sequence are combined to be used as training data for predicting the residual life of the power battery.
And (3) establishing a Gaussian mixture model to predict the residual life of the power battery by introducing multi-sequence discharge cut-off voltage data, and outputting a confidence interval.
2. The method for predicting the residual life of the similar point voltage sequence power battery as claimed in claim 1, wherein the problem that the capacity and internal resistance of the traditional direct health factor are difficult to realize on-line measurement is solved, and the indirect health factor is extracted: discharge cutoff time, average discharge current time, and discharge temperature peak time; grey correlation theory was introduced to determine the correlation between indirect health factors and power cell degradation characteristics. Considering the fact that the method is applicable to the actual vehicle-mounted working condition, the discharge voltage cut-off time is used as an indirect health factor for predicting the residual life of the power battery.
3. The discharge cut-off voltage time as recited in claims 1 and 2 is used as an indirect health factor, and the voltage sequence of each battery of the twin neural network is used for similarity quantitative analysis in consideration of error accumulation of single historical cycle data and failure to accurately reflect abnormal degradation information of the lithium ion battery. And introducing a voltage similarity theory, and representing the overall degradation characteristics of the battery and the unconventional degradation phenomenon by combining a monomer discharge voltage cut-off time sequence and a voltage similarity sequence.
4. The method as claimed in claims 1, 2 and 3, wherein under the actual complex vehicle-mounted working condition, obtaining the confidence coefficient of the remaining life of the lithium ion battery has very important practical value, and a Gaussian mixture regression remaining life prediction model is provided based on the monomer discharge cutoff voltage sequence and the similar voltage sequence. And selecting the first 50% of the degradation period as training data, and the last 50% as test data, and obtaining the residual life of the power battery and a confidence interval by adopting a Gaussian mixture model.
CN202110020952.XA 2021-01-07 2021-01-07 Power battery residual life indirect prediction method of similar voltage sequence Pending CN114740356A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110020952.XA CN114740356A (en) 2021-01-07 2021-01-07 Power battery residual life indirect prediction method of similar voltage sequence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110020952.XA CN114740356A (en) 2021-01-07 2021-01-07 Power battery residual life indirect prediction method of similar voltage sequence

Publications (1)

Publication Number Publication Date
CN114740356A true CN114740356A (en) 2022-07-12

Family

ID=82274063

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110020952.XA Pending CN114740356A (en) 2021-01-07 2021-01-07 Power battery residual life indirect prediction method of similar voltage sequence

Country Status (1)

Country Link
CN (1) CN114740356A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117124922A (en) * 2023-09-13 2023-11-28 广东一帆新能源有限公司 Power-off protection device and system for rechargeable battery of new energy automobile

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117124922A (en) * 2023-09-13 2023-11-28 广东一帆新能源有限公司 Power-off protection device and system for rechargeable battery of new energy automobile

Similar Documents

Publication Publication Date Title
Guo et al. Online estimation of SOH for lithium-ion battery based on SSA-Elman neural network
CN112731159B (en) Method for pre-judging and positioning battery faults of battery compartment of energy storage power station
CN110224192B (en) Method for predicting service life of power battery by gradient utilization
Xu et al. State-of-health estimation for lithium-ion batteries based on Wiener process with modeling the relaxation effect
Li et al. Fault diagnosis for lithium-ion batteries in electric vehicles based on signal decomposition and two-dimensional feature clustering
Song et al. Remaining life prediction of lithium-ion batteries based on health management: A review
CN111983474A (en) Lithium ion battery life prediction method and system based on capacity decline model
Zhou et al. Research on state of health prediction model for lithium batteries based on actual diverse data
CN116609676B (en) Method and system for monitoring state of hybrid energy storage battery based on big data processing
CN111460656B (en) Method and system for evaluating operation life of communication power supply of electric power machine room
CN115480180A (en) New energy battery health diagnosis and analysis method
CN114646888A (en) Assessment method and system for capacity attenuation of power battery
Shao et al. A novel method of discharge capacity prediction based on simplified electrochemical model-aging mechanism for lithium-ion batteries
Long et al. Multi-parameter optimization method for remaining useful life prediction of lithium-ion batteries
Ran et al. Fast clustering of retired lithium-ion batteries for secondary life with a two-step learning method
Liu et al. Review on degradation mechanism and health state estimation methods of lithium-ion batteries
Xu et al. Safety warning analysis for power battery packs in electric vehicles with running data
Wang et al. Dynamic early recognition of abnormal lithium-ion batteries before capacity drops using self-adaptive quantum clustering
CN114740356A (en) Power battery residual life indirect prediction method of similar voltage sequence
Huang et al. State of health estimation of lithium-ion batteries based on fine-tuning or rebuilding transfer learning strategies combined with new features mining
Huang et al. A novel label-free supervision learning method for lithium-ion battery RUL prediction
Lin et al. Li-ion battery state of health prediction based on long short-term memory recurrent neural network
Zhang et al. The early prediction of lithium-ion battery remaining useful life using a novel long short-term memory network
CN114579659A (en) System and method for estimating and sorting utilization potential of power battery by gradient utilization
CN115128468A (en) Chemical energy storage battery PHM undervoltage fault prediction method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
DD01 Delivery of document by public notice

Addressee: Wei Meng

Document name: Notification of Publication of Invention Patent Application

DD01 Delivery of document by public notice