Large-Scale Scenarios of Electric Vehicle Charging with a Data-Driven Model of Control

TL;DR

This paper introduces a fast, data-driven modeling approach for large-scale electric vehicle charging scenarios, incorporating control strategies to estimate demand accurately and efficiently for planning and rate design.

Contribution

A novel machine learning method for rapid, accurate modeling of controlled EV charging demand at large scales, replacing traditional optimization techniques.

Findings

Achieved 2.5% to 4.5% modeling errors.

Accelerated computations by over 4000 times.

Generated California 2030 charging scenarios in under 50 seconds.

Abstract

Planning to support widespread transportation electrification depends on detailed estimates for the electricity demand from electric vehicles in both uncontrolled and controlled or smart charging scenarios. We present a modeling approach to rapidly generate charging estimates that include control for large-scale scenarios with millions of individual drivers. We model uncontrolled charging demand using statistical representations of real charging sessions. We model the effect of load modulation control on aggregate charging profiles with a novel machine learning approach that replaces traditional optimization approaches. We demonstrate its performance modeling workplace charging control with multiple electricity rate schedules, achieving small errors (2.5% to 4.5%), while accelerating computations by more than 4000 times. We illustrate the methodology by generating scenarios for California's 2030 charging demand including multiple charging segments and controls, with scenarios run locally in under 50 seconds, and for assisting rate design modeling the large-scale impact of a new workplace charging rate.