CN-122017636-A - Diagnosis method for abnormal internal resistance of power battery

CN122017636ACN 122017636 ACN122017636 ACN 122017636ACN-122017636-A

Abstract

The invention relates to a power battery abnormal internal resistance diagnosis method, which comprises four steps of data acquisition and preprocessing, battery internal resistance estimation, normal internal resistance rule construction and abnormal battery internal resistance identification based on machine learning and data driving. The invention combines the machine learning method and the data driving method, adopts the machine learning method, combines a plurality of strategies such as the extraction of the driving fragment, the judgment of convergence, the normalization of error and the like to create the battery internal resistance estimation method, utilizes the data driving algorithm and the residual error estimation to diagnose the abnormal internal resistance of the battery, has better accuracy and authenticity and is more close to engineering application.

Inventors

LI DA
CHENG BOYUAN
LI WEILIN
WANG NINGHAO
REN YIFENG
LIU SONGYAN
LIU PEI
QI YANG

Assignees

西北工业大学

Dates

Publication Date: 20260512
Application Date: 20260206

Claims (6)

1. A power battery abnormal internal resistance diagnosis method comprises the following steps: Step one, data acquisition and preprocessing: extracting data related to battery fault diagnosis from automobile data, processing missing values and abnormal values, smoothing noise, and merging the data in a plurality of data sources; Estimating the internal resistance of the battery: the historical battery voltage and battery current are divided into y driving segments in time series, Filtering the single battery voltage and the battery current in the driving segment X by adopting a Savitzky-Golay filter, and then carrying out recursive calculation by using a recursive least square method to obtain the internal resistance of the driving segment; Step three, normal internal resistance rule construction: taking the battery aging state, the battery temperature and the battery SOC as inputs and the battery internal resistance value as output to construct a neural network model; step four, identifying abnormal internal resistance of the battery: And constructing a residual evaluation strategy based on the normal internal resistance area and based on different battery temperatures, driving mileage and normal internal resistance under the battery SOC, which are obtained by the neural network model, and evaluating and diagnosing abnormal internal resistance of the battery through residual.
2. The power battery abnormal internal resistance diagnosis method according to claim 1, wherein the data includes a battery voltage, a battery temperature, a battery current, a driving range, and a battery SOC.
3. The method for diagnosing abnormal internal resistance of a power battery according to claim 1, wherein in the second step, in the process of performing the recursive computation by the recursive least square method, the convergence of the internal resistance is additionally judged, and if the internal resistance is converged and the traveling segment is not yet ended, the internal resistance at that time is recorded and the subsequent traveling is divided into a new segment. In this way, a plurality of internal resistances can be obtained in one travel segment to enrich the internal resistance sample size.
4. The method for diagnosing abnormal internal resistance of a power battery according to claim 1, wherein in the third step, the neural network model is composed of 4 layers of an input layer, an RBF hidden layer, a full-connection hidden layer and an output layer. The internal resistance sample data is mapped to high dimension through the RBF hidden layer so as to enhance the separability of the internal resistance sample, further improve the MLP precision, and the normal internal resistance is output through nonlinear transformation of the fully-connected hidden layer.
5. The method for diagnosing abnormal internal resistance of a power battery according to claim 1, wherein in the third step, the input battery temperature is directly measured by a real vehicle battery temperature probe, the battery aging state is represented by accumulated driving mileage, the nonlinear relationship between the normal internal resistance and the input battery temperature, the driving mileage and the battery SOC is trained by using the normal vehicle history data, and a normal internal resistance curved surface is constructed.
6. The power battery abnormal internal resistance diagnosis method according to claim 1, wherein the specific steps of the fourth step are: The residual between the actual internal resistance and the normal internal resistance can be expressed as: In the formula, Is residual; the internal resistance of the driving segment calculated in the second step is calculated; The normal internal resistance fitted for the neural network, T is the battery temperature, m is the driving mileage, S is the battery SOC; the normal internal resistance boundary is determined by the internal resistance residual error of the normal battery, and is expressed as: In the formula, Is a normal battery Average value of (2); Is a normal battery Is a variance of (2); Is a normal internal resistance boundary, is obtained by statistics of a normal battery, and is 0.066; The normal internal resistance region is: at the same temperature, mileage and SOC, the residual error between the two was evaluated by the following criteria: if the internal resistance of the battery meets the above criteria, the internal resistance of the battery cell exceeds the normal internal resistance region, and is diagnosed as an abnormal internal resistance battery.

Description

Diagnosis method for abnormal internal resistance of power battery Technical Field The invention belongs to the technical field of electrical engineering, and particularly relates to a method for diagnosing abnormal internal resistance of a power battery. Background The power battery is an energy source of the electric automobile and is also the part most prone to faults, and along with the continuous increase of the usage amount of the lithium ion battery, the safety problems such as thermal runaway and the like of the battery are more and more increased. The internal resistance is one of important parameters reflecting the safety of the battery, and the battery with excessive internal resistance can generate excessive heat, so that the battery is easier to cross a thermal runaway critical condition to generate irreversible thermal runaway. The accurate abnormal internal resistance diagnosis can discover potential safety hazards in advance, and thermal runaway is avoided. The existing battery internal resistance estimation method is based on high-frequency voltage and current under stable working conditions such as a laboratory pulse cycle test or a dynamic stress test, has low precision under sparse parameter and random current working conditions, and cannot effectively diagnose abnormal internal resistance of a battery under a real-vehicle scene. The power battery is an energy source of the electric automobile and is also the part most prone to faults, and along with the continuous increase of the usage amount of the lithium ion battery, the safety problems such as thermal runaway of the battery are more and more, and according to statistics of China automobile technical research center, the new energy automobile has 50 safety accidents from 1 month to 9 months in 2018, the safety accidents relate to casualties of a large number of people, and the problem to be solved most urgently in the battery development process is solved. The internal resistance is one of important parameters reflecting the safety of the battery, and the battery with excessive internal resistance can generate excessive heat, so that the battery is easier to cross a thermal runaway critical condition to generate irreversible thermal runaway. The accurate abnormal internal resistance diagnosis can discover potential safety hazards in advance, and thermal runaway is avoided. The existing battery internal resistance estimation method is based on high-frequency voltage and current under stable working conditions such as a laboratory pulse cycle test or a dynamic stress test, has low precision under sparse parameter and random current working conditions, and cannot effectively diagnose abnormal internal resistance of a battery under a real-vehicle scene. Disclosure of Invention Aiming at the current state of the art, the invention aims to solve the technical problems that the data of the power battery in the running process of the new energy automobile is analyzed through machine learning and a data driving method, the abnormal internal resistance of the battery is diagnosed, and the occurrence of battery safety accidents is reduced. The method can diagnose the abnormal internal resistance of the battery with high accuracy under the actual vehicle scene of sparse parameters and random current working conditions. A power battery abnormal internal resistance diagnosis method comprises the following steps: Step one, data acquisition and preprocessing: extracting data related to battery fault diagnosis from automobile data, processing missing values and abnormal values, smoothing noise, and merging the data from a plurality of data sources. The data includes battery voltage, battery temperature, battery current, mileage, battery SOC. Estimating the internal resistance of the battery: the historical battery voltage and battery current are divided into y driving segments in time series, And filtering the single battery voltage and the battery current in the driving segment X by adopting a Savitzky-Golay filter, and then carrying out recursive calculation by using a recursive least square method to obtain the internal resistance of the driving segment. Step three, normal internal resistance rule construction: And taking the battery aging state, the battery temperature and the battery SOC as inputs and the battery internal resistance value as output to construct the neural network model. Step four, identifying abnormal internal resistance of the battery: And constructing a residual evaluation strategy based on the normal internal resistance area and based on different battery temperatures, driving mileage and normal internal resistance under the battery SOC, which are obtained by the neural network model, and evaluating and diagnosing abnormal internal resistance of the battery through residual. In the second step, in the process of performing the recursive computation by the recursive least square method, the convergence of the internal resistan