Abstract:Neural networks are proving to be a valuable tool in the design and application of batteries for electric vehicles. They are being employed because of their ability to generate accurate battery models using only experimental data. Not only are these models simpler than conventional ones based on explicit math, they also consider more design parameters.

Neural-network battery models account for cell size, geometry, mass, and rate of discharge. From this, they can estimate battery output in terms of energy density, power density, efficiency, and service life. Given sufficient experimental data, the models can interpolate (and sometimes even extrapolate) battery performance over a wide range of conditions. And they work for any technology, including lead acid, sodium sulfur, nickel metal hydride, and zinc bromide.