Impact of an electrified parkade on the built environment: An unsupervised learning approach

Abstract

Proposed legislation in British Columbia would require 30 percent of new car sales to be zero-emission vehicles by 2030, and 100 percent by 2040. The growing amount of energy demand and usage data from smart meters or internet of things (IoT) devices enables new research areas. We reporton machine learning approaches to reevaluate the impacts of battery electric vehicles (BEV) on the built environment. We developed a daily power profile analysis based on unsupervised learning, to understand the underlying structure of building and BEV charging station demand data. In addition, we have implemented a load aggregation method based on the features revealed by a clustering process. This aggregation method simulates the electricity demand of an arbitrary number of charging stations, all of which are connected to the main feeder of a building. Several scenarios are simulated using charging stations and building demand data from the University of British Columbia campus in Vancouver. Results for 150 charging stations revealed that the feeder load could increase from a peak load scenario of 300 kW to more than 1000 kW during a high-consumption weekday.