I-Corps: Integrating Electric Vehicles in the Smart Grid Through Smart Charging Software

University Of California-Berkeley, Berkeley CA

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is to enable safe integration of Plug-in Electric Vehicles (PEVs) into the broader power network, as well as to reduce PEV charging costs for drivers. If unmanaged, PEVs would represent major extra loads for utilities, which would increase energy peak consumption and require considerable investment to upgrade grid infrastructure. The technology developed here involves a smart charging software, which will enable real time communication between PEV fleets and utilities, and will remotely control PEV charging to provide grid regulation services. This project will bridge the gap between the automotive industry and the energy industry, and will develop strategies to create a smart and connected ecosystem around PEV charging. Successful commercialization of the smart charging software could result in lower PEV cost of ownership, and increased penetration of clean transportation in the US. This I-Corps project is based on an advanced, adaptive, optimization framework for Plug-in Electric Vehicle (PEV) charging control in power networks. The technique can control PEV charging under conditions of general uncertainty, and consider both driver mobility constraints and power network constraints. The optimization framework is particularly adapted to large fleets of PEVs and shows fast convergence rate and privacy preserving properties. The method uses past PEV data to predict travel behaviors and energy demand. It also models distribution grid constraints and electricity market structure to set relevant optimization objectives. The method utilizes a range of optimization methods and selects the best algorithm based on tradeoff between optimality, constraint satisfaction and computation time. The methods include partial differential equation aggregation techniques for car sharing fleets, plug-and-play model predictive control for power networks with high congestion, and dual-splitting methods for large scale residential fleets with lower computation capabilities.