The EV equivalent of a gas gauge—the range estimator—is notoriously imprecise, so much so that some drivers call it a “guess-o-meter” or “guess gauge.” Obviously, the remaining range of a particular EV on a particular day depends on many factors (speed, traffic conditions, weather). However, it turns out that even estimating a battery’s state of charge is not so simple.

Electra Vehicles, a provider of predictive battery management and battery design software, has demonstrated a new system designed to improve the accuracy of EV driving range estimates. Electra says its core technology—EVE-Ai Adaptive Cell Modeling System—outperformed the industry standard for estimating battery charge, resulting in a twofold reduction in estimation error.

Electra partnered with a semiconductor provider to construct a battery pack that was capable of delivering real-time battery cell data from the pack to Electra’s cloud-based EVE-Ai software through a battery management system and IoT gateway hardware. Using this setup, Electra found that its integrated software solution could retrain the battery management system using machine learning to predict a battery’s state of charge more accurately than the industry standard method, known as Extended Kalman Filtering (EKF).

The test battery pack was repeatedly charged and discharged over a 12-week period in order to quickly age the pack to roughly half of its warranty lifetime. Throughout the testing, Electra compared three sets of results: estimates from Electra’s EVE-Ai Adaptive Cell Modeling System; estimates from the industry standard EKF; and the reference values from an electrochemical reference data set.

The results showed that Electra’s solution better predicted the battery’s state of charge at the beginning of life, but more importantly, as the battery reached half-life, Electra’s accuracy improved significantly over EKF.

“By showcasing significant improvements in predicting a battery’s state-of-charge, Electra has demonstrated how using artificial intelligence in battery management can translate to longer-lasting and better-performing batteries,” said Electra CEO and co-founder Fabrizio Martini. “With Electra’s EVE-Ai software, the vehicle’s battery management system is constantly retrained to showcase the most accurate battery metrics.”

Source: Electra Vehicles