Statistical Modeling for Prognostics and Health Management of Rechargeable Batteries

用於充電電池退化預測和健康管理的統計模型

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
  • Kwok Leung TSUI (Supervisor)
Award date29 Jun 2020

Abstract

Rechargeable batteries have become a vital component in our daily lives and their applications have increased immensely from mobile phones and laptops in the 1990s to smartwatches, tablets, electric vehicles, and energy storage systems in recent times. This thesis revolves around the prognostics and health management (PHM) techniques on rechargeable batteries, where PHM has been a hot research topic in recent decades for monitoring and predicting the performance of a rechargeable battery. A key point of interest in this thesis is to use statistical modeling method and processed data to identify and monitor the health status of rechargeable batteries, predict the degradation and prevent system failure associated with battery degradation. The contributions of this thesis are as follows:

• To propose a robust cubic smoothing spline method and peak height ratio indicator for the state of health (SOH) estimation and state of charge (SOC) recalibration through incremental capacity analysis

To improve the efficiency of SOH estimation, incremental capacity analysis is one of the methods that allows estimation of the SOH without the need of full charge or discharge profile. A robust cubic smoothing spline method is proposed to obtain the incremental capacity (IC) curve, with a main advantage that the smoothing parameter can be determined using cross validation instead of subjective trial and error parameter tuning appeared in typical filtering methods. Comparison through simulated data shows that the robust cubic smoothing spline has shown good fits on data and feature of interest with low root mean square error (RMSE) values. Besides, a peak height ratio feature extracted within the IC curve is proposed for estimating the SOH through the linear relationship observed between SOH and peak height ratio. A more generalized version of the SOH estimation method, which is expected to be more applicable to different degradation mechanisms, is also demonstrated using a multiple linear regression with covariates of both peak height ratio and the height of peak associated with “last stage of Li ions intercalation during charging”. The application of the SOH estimation method is demonstrated with lithium iron phosphate (LFP) battery cycle aging data. Also, the method can be applied to update the maximum available capacity for recalibrating the SOC estimation.

• To develop a time series model for degradation modeling and remaining useful life (RUL) prediction

While most existing methods consider a deterministic degradation model such as an exponential model, we have developed a semi-parametric degradation model based on a multiple-change-point linear model with two sub-models: an autoregressive model with covariates and a survival regression model. The proposed model is capable of predicting battery degradation path at the initial stage of battery life based on historical paths, predicting the remaining degradation path based merely on the partially observed path through constructing segments iteratively with autoregressive nature, producing confidence intervals (CIs) for the battery life from a bootstrap distribution, as well as predicting battery degradation with temperature influence incorporated. The model has been applied to some experimental and simulated degradation data from two different battery types, LFP and lithium nickel manganese cobalt oxide (NMC) batteries. The validation with simulation shows that the proposed model is reliable when complete historical paths are available as the simulation coverage rates are close to the nominal coverage rate of 90%. The proposed model is intended to provide an alternative perspective for examining battery degradation data instead of a deterministic model.

• To develop a two-stage gamma process model with a fixed change point for prognostics

We have developed a two-stage gamma process model with a fixed change point for modeling the discharge voltage curve, which can be subsequently applied to estimate the SOC and remaining useful discharge time (RUT) in a cycle while taking into account the effect of cycle aging, as well as to estimate the state of life (SOL) and RUL, under constant current rate scenario. The model is able to produce CIs of such estimations using a parametric bootstrap method. The applications of the proposed model are demonstrated using the experimental cycle aging data of LFP batteries. We have also compared the proposed model with the two-term exponential model under a particle filter framework on the battery experimental cycle aging data. The result shows that the proposed model can obtain a more accurate prediction than the conventional method in which only capacity-fading data is modeled. This reveals the superiority of modeling with voltage data over modeling with capacity data. In addition, we have developed an analytical expression of mean useful discharge time in a cycle (or mean time to failure) which is approximated by a Taylor expansion and the Birnbaum-Saunders distribution, and the result is shown to be in good agreement with the true mean of a gamma process.