Machine-Learning-Based Efficient Parameter Space Exploration for Energy Storage Systems

Abstract

Shifting towards sustainable energy sources requires developing new storage systems and estimating their durability. Durability depends on many operating parameters, resulting in a large and high-dimensional parameter space that should be explored. Testing cells exhaustively on a dense grid in the parameter space is prohibitively expensive. This is especially true with considerable cell-to-cell variability in performance, even under same cycling conditions. Here, we develop a framework based on Gaussian processes, equipped with domain knowledge, to implement a Bayesian optimization approach to explore the parameter space efficiently and quantify durability using failure probability distributions. Bayesian optimization identifies future experiments that maximize information gain and minimize uncertainty. Experimental results show accurate durability predictions with a significantly reduced number of experiments. However, laboratory cycling conditions, including those in the literature, may not represent real-world cycling. For this, we propose an approach based on laboratory results to predict durability under real-world cycling conditions accurately.

Keywords: Machine Learning, Bayesian Optimization, Gaussian Process, Stochastic Modeling, Parameter Space, Efficient Exploration, Energy Storage, Battery Failure Modeling, Complex Cycling Conditions