arXiv subjects

Antti Aitio

Publications and source records attributed to Antti Aitio.

4 recordsLinked to original sources

arXiv2025

Learning Li-ion battery health and degradation modes from data with aging-aware circuit models

Non-invasive estimation of Li-ion battery state-of-health from operational data is valuable for battery applications, but remains challenging. Pure model-based methods may suffer from inaccuracy and long-term instability of parameter estimates, whereas pure data-driven methods rely heavily on training data quality and quantity, causing lack of generality when extrapolating to unseen cases. We apply an aging-aware equivalent circuit model for health estimation, combining the flexibility of data-driven techniques within a model-based approach. A simplified electrical model with voltage source and resistor incorporates Gaussian process regression to learn capacity fade over time and also the dependence of resistance on operating conditions and time. The approach was validated against two datasets and shown to give accurate performance with less than 1% relative root mean square error (RMSE) in capacity and less than 2% mean absolute percentage error (MAPE). Critically, we show that the open circuit voltage versus state-of-charge function must be accurately known, and any inaccuracies or changes in this over time strongly influence the inferred resistance. However, this feature (or bug) may also be used to estimate in operando differential voltage curves from operational data.

eess.SY↗

arXiv2023

Learning battery model parameter dynamics from data with recursive Gaussian process regression

Estimating state of health is a critical function of a battery management system but remains challenging due to the variability of operating conditions and usage requirements of real applications. As a result, techniques based on fitting equivalent circuit models may exhibit inaccuracy at extremes of performance and over long-term ageing, or instability of parameter estimates. Pure data-driven techniques, on the other hand, suffer from lack of generality beyond their training dataset. In this paper, we propose a hybrid approach combining data- and model-driven techniques for battery health estimation. Specifically, we demonstrate a Bayesian data-driven method, Gaussian process regression, to estimate model parameters as functions of states, operating conditions, and lifetime. Computational efficiency is ensured through a recursive approach yielding a unified joint state-parameter estimator that learns parameter dynamics from data and is robust to gaps and varying operating conditions. Results show the efficacy of the method, on both simulated and measured data, including accurate estimates and forecasts of battery capacity and internal resistance. This opens up new opportunities to understand battery ageing in real applications.

eess.SY↗

arXiv2021

Predicting battery end of life from solar off-grid system field data using machine learning

Hundreds of millions of people lack access to electricity. Decentralised solar-battery systems are key for addressing this whilst avoiding carbon emissions and air pollution, but are hindered by relatively high costs and rural locations that inhibit timely preventative maintenance. Accurate diagnosis of battery health and prediction of end of life from operational data improves user experience and reduces costs. But lack of controlled validation tests and variable data quality mean existing lab-based techniques fail to work. We apply a scaleable probabilistic machine learning approach to diagnose health in 1027 solar-connected lead-acid batteries, each running for 400-760 days, totalling 620 million data rows. We demonstrate 73% accurate prediction of end of life, eight weeks in advance, rising to 82% at the point of failure. This work highlights the opportunity to estimate health from existing measurements using `big data' techniques, without additional equipment, extending lifetime and improving performance in real-world applications.

cs.LG↗

arXiv2020

Bayesian Parameter Estimation Applied to the Li-ion Battery Single Particle Model with Electrolyte Dynamics

This paper presents a Bayesian parameter estimation approach and identifiability analysis for a lithium-ion battery model, to determine the uniqueness, evaluate the sensitivity and quantify the uncertainty of a subset of the model parameters. The analysis was based on the single particle model with electrolyte dynamics, rigorously derived from the Doyle-Fuller-Newman model using asymptotic analysis including electrode-average terms. The Bayesian approach allows complex target distributions to be estimated, which enables a global analysis of the parameter space. The analysis focuses on the identification problem (i) locally, under a set of discrete quasi-steady states of charge, and in comparison (ii) globally with a continuous excursion of state of charge. The performance of the methodology was evaluated using synthetic data from multiple numerical simulations under diverse types of current excitation. We show that various diffusivities as well as the transference number may be estimated with small variances in the global case, but with much larger uncertainty in the local estimation case. This also has significant implications for estimation where parameters might vary as a function of state of charge or other latent variables.

eess.SY↗