A Two-Layers Predictive Algorithm for Workplace EV Charging

TL;DR

This paper presents a two-layer predictive algorithm for workplace EV charging that optimizes costs and schedules charging based on demand predictions and dynamic programming, validated with real data.

Contribution

It introduces a novel two-layer approach combining demand prediction and scheduling to improve EV charging efficiency and cost savings at workplaces.

Findings

The proposed scheme reduces charging costs compared to conventional methods.

It effectively manages EV load demand without requiring user departure time input.

Validation with real-world data demonstrates practical applicability.

Abstract

In this paper, the problem of electric vehicle (EV) charging at the workplace is addressed via a two-layer predictive algorithm. We consider a time of use (TOU) pricing model for energy drawn from the grid and try to minimize the charging cost incurred by the EV charging station (EVCS) operator via an economic layer based on dynamic programming (DP) approach. An adaptive prediction algorithm based on a non-parametric stochastic model computes the projected EV load demand over the day which helps in the selection of optimal loading policy for the EVs in the economic layer. The second layer is a scheduling algorithm designed to share the allocated power limit (obtained from economic layer) among the charging EVs during each charge cycle. The modeling and validation is performed using ACN data-set from Caltech. Comparison of the proposed scheme with a conventional DP algorithm illustrates its effectiveness in terms of supplying the requested energy despite lacking user input for departure time.