Virtual DavidAI

The work in brief

David's work addresses how to model, monitor and manage batteries from the single cell to grid-scale systems. Early papers (2012–2016) focus on battery management: fast solution of the Doyle–Fuller–Newman model by spectral orthogonal collocation, Kalman-filter state estimation, observability and structural identifiability of equivalent-circuit and fractional-order models, and sensorless internal temperature estimation from impedance. A recurring methodological thread is the question of what can actually be learned from data. It appears in identifiability analysis of the single particle model and in Bayesian parameter estimation with MCMC. Later it extends to Bayesian model selection via Bayesian quadrature, the PyBOP parameterisation package, and impedance-based parametrisation of physics-based models (PyBaMM-EIS, nonlinear and multisine EIS).

From around 2017 the work increasingly applies probabilistic machine learning to battery health. Gaussian process regression is used for capacity estimation and state-of-health forecasting, and automated feature selection for knee-point and end-of-life prediction. Hierarchical Bayesian models address lifetime prediction and cell-to-cell variability, and aging-aware circuit models estimate degradation modes from operational field data, including over a thousand off-grid solar lead-acid batteries. In parallel the scope widens to systems and applications: degradation-aware optimal dispatch and service stacking for grid batteries, a digital twin of a MWh-scale system, current and temperature imbalance and fault detection in parallel-connected packs, and electrochemical–thermal characterisation of large pouch cells. Further papers cover off-grid energy access and community efforts on battery data standards, linked data and verifiable modelling.

AI-written from the paper titles and abstracts; David hasn't reviewed it.