Title: The Role of Driver Behavior in Supporting Renewables in the Electric Grid
Abstract: As we move towards more renewable resources, the ability to produce electricity in time with demand diminishes. Instead, rises a need for energy storage or the ability to produce electricity when renewables allow and store it for when demand needs it later. Electric Vehicles (EVs) have been discussed as a way of providing a distributed energy storage resource to the electric grid. However, before EVs can be used to support the grid, through optimal charging and discharging, highly complex, non-linear driver behavior must be accounted for. In this work we closely collaborate with a large American EV manufacturer to combine machine learning with optimization to model driver behavior in order to size the capacity of energy storage EVs can offer to the grid. We demonstrate driver behavior can be forecasted with a random forest method. This motivates us to introduce a novel upper-bounding method for optimizing over tree ensemble models tractably. Using this method, we optimize the battery charging/discharging policy of each driver over the course of two weeks. We prove analytically and computationally that this new method performs strongly both in terms of runtime and analytical optimality gap. Furthermore, through our industry collaboration, we show that optimal charging and discharging of the car’s battery can result in $1,069 saved and carbon benefit equivalent to 77.3% of a home’s energy consumption annually. Our overarching goal though this work is to further align machine learning and optimization in service of sustainability.
Bio: Leann Thayaparan is a fifth-year PhD student at the MIT, Operations Research Center where she is being advised by Prof. Georgia Perakis. Her research focuses on optimization problems learned from data-driven machine learning, with a focus on sustainable operations. Her research has been in collaboration with companies such as General Motors, Oracle Retail, the CDC and MIT Quest for Intelligence. Before joining the PhD program, Leann received her Master’s of Business Analytics (2019) at MIT, worked at McKinsey and Co in their Operations Advance Analytics group and graduated with high honors in Operations Research and Financial Engineering (2016) from Princeton University.
