Dissertation/Thesis Abstract

Lithium-Ion Battery Degradation Evaluation through Bayesian Network Method for Residential Energy Storage Systems
by Khan, Khalid, Ph.D., Michigan Technological University, 2019, 172; 27668394
Abstract (Summary)

Batteries continue to infiltrate in innovative applications with the technological advancements led by Li-ion chemistry in the past decade. Residential energy storage is one such example, made possible by increasing efficiency and decreasing the cost of solar PV. Residential energy storage, charged by rooftop solar PV is tied to the grid, provides household loads. This multi-operation role has a significant effect on battery degradation. These contributing factors especially solar irradiation and weather conditions are highly variable and can only be explained with probabilistic analysis. However, the effect of such external factors on battery degradation is approached in recent literature with mostly deterministic and some limited stochastic processes. Thus, a probabilistic degradation analysis of Li-ion batteries in residential energy storage is required to evaluate aging and relate to the external causal factors. The literature review revealed modified Arrhenius degradation model for Li-ion battery cells. Though originating from an empirical deterministic method, the modified Arrhenius equation relates battery degradation with all the major properties, i.e. state of charge, C-rate, temperature, and total amp-hour throughput.

These battery properties are correlated with external factors while evaluation of capacity fade of residential Li-ion battery using a proposed detailed hierarchical Bayesian Network (BN), a hierarchical probabilistic framework suitable to analyze battery degradation stochastically. The BN is developed considering all the uncertainties of the process including, solar irradiance, grid services, weather conditions, and EV schedule. It also includes hidden intermediate variables such as battery power and power generated by solar PV. Markov Chain Monte-Carlo analysis with Metropolis-Hastings algorithm is used to estimate capacity fade along with several other interesting posterior probability distributions from the BN. Various informative and promising results were obtained from multiple case scenarios that were developed to explore the effect of the aforementioned external factors on the battery. Furthermore, the methodologies involved to perform several characterizations and aging test that is essential to evaluate the estimation proposed by the hierarchical BN is explored. These experiments were conducted with conventional and low-cost hardware-in-the-loop systems that were developed and utilized to quantify the quality of estimation of degradation.

Indexing (document details)
Advisor: Gauchia, Lucia, Ten, Chee-Wooi
Commitee: Roggemann, Michael C., Weaver, Wayne W., Brown, Laura E.
School: Michigan Technological University
Department: Electrical & Computer Engineering
School Location: United States -- Michigan
Source: DAI-B 81/7(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Electrical engineering, Engineering
Keywords: Bayesian models, Energy storage systems, Li-ion batteries, Probabilistic analysis, Residential energy storage system, Smart home
Publication Number: 27668394
ISBN: 9781392689875
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest