Lithium-ion battery SOH introduction International lithium-ion battery SOH estimation method

Based on the research on the estimation method of battery SOH at home and abroad, this paper introduces two internationally accepted mainstream methods.

1. Experience-based approach

The empirical battery life prediction method is also called the statistical rule based method, which mainly includes the following three types:

1) Cycle number method

This method is to count the cycle of the battery, and when the number of cycles of the battery reaches a certain range, it is considered that the battery reaches the service life. This method needs to consider the effects of different cycling conditions, cycle states and other factors on cycle life, and determine battery life based on both experience and standard parameters.

2) Anthony method and weighted ampere-time method

The total number of hours that a battery can handle power during the whole process of charging and discharging from a new battery should be a fixed value when the accumulated battery capacity reaches a certain level, the battery is considered to have reached the end of its life. This method is an ampere-hour method. The weighted ampere-hour method considers that when the battery emits the same amount of electricity under different conditions, the degree of damage to the life is light and heavy, so when the accumulated amount of electricity is multiplied by a weighting factor, the accumulated ampere-hours reaches a certain value and the battery is considered to arrive. End of life.

3) Event-oriented aging accumulation method

This method first needs to formulate a description of the specific event that causes the loss of battery life. Generally, each event has a scale description of the degree of damage, monitoring the occurrence of events during the use of the battery, and accumulating the battery life decay caused by each event, the remaining life of the current battery.

The above methods are all based on some experience in the use of batteries, according to some statistical laws to give a rough estimate of battery life, only in the case of sufficient experience of battery use, for specific occasions Life prediction.

2, performance-based approach

Based on various forms of performance models, and considering the aging process and stress factors. At present, many studies have carried out battery life based life prediction based on this idea. According to the different sources of information used in life prediction, life prediction based on battery performance is divided into three categories: mechanism-based, feature-based, and data-driven.

The mechanism-based prediction is to analyze and establish the operating mechanism model and aging model of the battery from the perspective of the essential mechanism of the battery, describe the aging behavior of the battery from the perspective of electrochemical principle, and predict the battery life by analyzing the battery model.

Feature-based prediction is the evolution of the characteristic parameters exhibited in the aging process of the battery, and the correspondence between the feature quantity and the battery life is established for life prediction.

Data-driven prediction is the use of battery performance test data, mining the law of battery performance evolution from the data for life prediction. For example, analytical models and artificial neural network models derived from data fitting are data-driven methods. Each of the three methods has its own advantages and disadvantages, and the combination of several methods is often used in practical applications.

1) Mechanism-based approach

Mechanism-based prediction requires studying the effect of each aging factor on state variables. This method first describes the physicochemical process of the battery, based on Ohm's law, Kirchhoff's voltage-current law, and electrochemical reaction process (Butler-Volmer). Law), diffusion process (Fick's law), etc.; then study the law of the influence of the aging process on state variables. On the one hand, we should study the mechanism model of the battery on the other hand. We should study the aging mechanism model of the aging process and the influence of stress factors on the state variables.

The main advantages of mechanism-based life prediction are: batteries suitable for almost all state conditions and operating modes; a detailed explanation of the battery aging process, which can be used for battery production and design manufacturers to improve battery design; compared with other methods Based on the model, the analysis of the battery control strategy can be more detailed and accurate. The disadvantage is that the model requires fine parameters and high complexity; the test for aging factors is complicated, and it is difficult to establish a perfect aging mechanism model.

2) Feature-based prediction method

The idea based on feature life prediction is to use the evolution of the characteristic parameters exhibited in the battery aging process to establish the correspondence between the feature quantity value and the battery health status for life prediction.

Current feature-based battery life predictions are primarily focused on the relationship between electrochemical impedance and battery cycle life. Electrochemical impedance spectroscopy (EIS) is used as the research method of battery life characteristics. Generally, the impedance spectrum curve is measured at different stages of battery cycle life. The battery equivalent circuit model form is obtained according to the impedance spectrum curve, and the cycle number and etc. are analyzed. The influence law of the parameters such as solution resistance, load resistance and Warburg impedance in the effective circuit model, and finally the fitting formula of the parameters in the equivalent circuit model with the number of battery cycles is given. In addition to the EIS impedance spectrum, there is also a pulse impedance measurement method that estimates the internal resistance of a pulse or step excitation signal applied to the battery.

The EIS impedance spectrum can give a more detailed description of the battery impedance and can be used to estimate the life characteristics of the battery. However, the measurement is more complicated and requires special measuring instruments. The EIS technology is applied to the on-line monitoring of the battery state. Rapid measurement techniques are being studied. Impulse impedance measurement is simple and easy to perform. It can be measured quickly and can be monitored online. The test results can describe the impedance of the battery to a certain extent, reflect the characteristics of the battery impedance increasing with the decay of life, and can also be used as the battery life characteristics.

3) Data-driven prediction

The physicochemical process of the battery itself is complicated, and many laws are difficult to describe directly through mechanism research. The idea of describing battery performance from the perspective of test data is called a data-driven approach.

There are many common data-driven algorithms, such as Support Vector Machine (SVM), Autoregressive Moving Average (ARMA), and Particle Filtering (PF).

Data-driven prediction does not require the mechanism knowledge of the object system. Based on the collected data, various data analysis learning methods are used to mine the implicit information for prediction, thus avoiding the complexity of model acquisition, practical forecasting method. However, the data obtained usually tends to have strong uncertainty and incompleteness. It is unrealistic to test all the possible life-influencing factors in practical applications. Therefore, data-driven predictions are easy to implement, but they also have certain limitations.

The page contains the contents of the machine translation.