Nowadays, batteries in electric vehicles (EVs) are facing a variety of tasks in their connection to the power grid in addition to the main task, driving. All of these tasks play a very significant role in the battery aging, but they are highly variable due to the change in the driver behavior, grid connection availability and weather conditions. The effect of these external factors in the battery degradation have been studied in literature by mostly deterministic and some stochastic approaches, but limited to specific cases.

In this dissertation, first, a large-scale deterministic approach is implemented to evaluate the effect of variations in the EV battery daily tasks. To do so, a software tool named REV-Cycle is developed to simulate the EV powertrain and studied the effect of driving behavior, recharging facilities and timings, grid services and temperature/weather change effects, one by one. However, there are two main problems observed in the deterministic aging evaluation: First, the battery capacity fade factors such as temperature, cycling current, state of charge (SOC)...are dependent to the external variables such as location, vehicle owner's behavior and availability of the grid connection. Therefore, it is not possible to accurately evaluate the battery degradation with a deterministic model, while its inputs are stochastic. Second, the battery aging factors' dependency is hierarchical and it is not easy to follow and implement this hierarchy with deterministic models.

Therefore, using a hierarchical probabilistic framework is proposed that can better represent the problem and realized that the Bayesian statistics with Markov Chain Monte Carlo (MCMC) can provide the problem solving structure needed for this purpose. A comprehensive hierarchical probabilistic model of the battery capacity fade is proposed using Hierarchical Bayesian Networks (HBN). The model considers all uncertainties of the process including vehicle acceleration and velocity, grid connection for charging and utility services, temperatures and all unseen intermediate variables such as battery power, auxiliary power, efficiencies, etc. and estimates the capacity fade as a probability distribution. Metropolis-Hastings MCMC algorithm is applied to generate the posterior distributions. This modeling approach shows promising result in different case studies and provides more informative evaluation of the battery capacity fade.