A team of scientists have used artificial learning to predict electric vehicle battery states to give an ‘accurate’ prediction for lithium-ion cells state of charge and health.
The data-driven machine learning model technology could enable manufacturers to embed the software straight into their battery devices to improve its cycle life up to 6% over typical battery models that miscalculate lifetimes by around 10%.
The machine learning software is ready to bring to market, but the hardware needs to be created to deploy the software on devices, which could take two years to redesign hardware to include standard computer chips.
The technology was developed during by the UK’s University of Cambridge, the Institute of Materials Research and Engineering at The Agency for Science, Technology and Research (A*STAR), and Nanyang Technological University in Singapore.
The review article was published in Nature Machine Intelligence.
The three long-standing problems for battery state prediction, where machine learning has the potential to make significant inroads, includes: holistic battery modelling that transcends time, length, and mechanism scales; accelerating and simplifying calculations to enable them to be done in-situ on the battery itself; and high-throughput computational and experimental data generation.
The article concluded: “Currently, the two most studied models for battery state prediction are the equivalent circuit model (ECM) and physics-based model (PBM).
“Despite their popularity and continuous development, there remains a clear tradeoff between computational efficiency and accuracy when using these models for on-line battery state prediction.
“Data driven models with machine learning is a promising way to model batteries that can potentially address the dilemma faced by traditional modelling using ECMs or PBMs.
“Currently, most of the machine learning models give ‘black box’ battery state predictions, which makes it difficult to generalise to other battery chemistries.”