Prognosis of Multivariate Battery State of Performance and Health via Transformers

by   Noah H. Paulson, et al.

Batteries are an essential component in a deeply decarbonized future. Understanding battery performance and "useful life" as a function of design and use is of paramount importance to accelerating adoption. Historically, battery state of health (SOH) was summarized by a single parameter, the fraction of a battery's capacity relative to its initial state. A more useful approach, however, is a comprehensive characterization of its state and complexities, using an interrelated set of descriptors including capacity, energy, ionic and electronic impedances, open circuit voltages, and microstructure metrics. Indeed, predicting across an extensive suite of properties as a function of battery use is a "holy grail" of battery science; it can provide unprecedented insights toward the design of better batteries with reduced experimental effort, and de-risking energy storage investments that are necessary to meet CO2 reduction targets. In this work, we present a first step in that direction via deep transformer networks for the prediction of 28 battery state of health descriptors using two cycling datasets representing six lithium-ion cathode chemistries (LFP, NMC111, NMC532, NMC622, HE5050, and 5Vspinel), multiple electrolyte/anode compositions, and different charge-discharge scenarios. The accuracy of these predictions versus battery life (with an unprecedented mean absolute error of 19 cycles in predicting end of life for an LFP fast-charging dataset) illustrates the promise of deep learning towards providing deeper understanding and control of battery health.


Transfer Learning and Vision Transformer based State-of-Health prediction of Lithium-Ion Batteries

In recent years, significant progress has been made in transportation el...

Attention-based Deep Neural Networks for Battery Discharge Capacity Forecasting

Battery discharge capacity forecasting is critically essential for the a...

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...

To Charge or To Sell? EV Pack Useful Life Estimation via LSTMs and Autoencoders

Electric Vehicles (EVs) are spreading fast as they promise to provide be...

Towards Integrating Formal Verification of Autonomous Robots with Battery Prognostics and Health Management

The battery is a key component of autonomous robots. Its performance lim...

Cerberus: A Deep Learning Hybrid Model for Lithium-Ion Battery Aging Estimation and Prediction Based on Relaxation Voltage Curves

The degradation process of lithium-ion batteries is intricately linked t...

Improve in-situ life prediction and classification performance by capturing both the present state and evolution rate of battery aging

This study develops a methodology by capturing both the battery aging st...

Please sign up or login with your details

Forgot password? Click here to reset