Session: DAC-04-02: Data-Driven Design

Paper Number: 116751

116751 - Using Machine Learning to Predict the Adoption of Building Electrification Technologies in Us Households 

This paper explores the use of machine learning techniques to predict the adoption of building electrification technologies within US households. This is important due to the increasing prevalence of building electrification as a pathway to addressing climate change, which inadvertently poses a threat to the energy resilience of households during power outages. A non-intrusive, data-driven means of predicting the level of technology adoption can help in the deign of mitigation and adaptation strategies aimed at minimizing the risks vulnerable households may face when power outages are compounded by extreme weather events. This study develops machine learning models based on the energy consumption dynamics of US households to predict the presence of critical electric appliances, including electric furnace, electric water heater, induction stove, electric cooling systems, and solar panels. The models are trained using a large dataset of building end-use load consumption for buildings located in New Jersey, and sourced from the US Department of Energy. The results of this work show that the machine learning models are reasonably accurate in predicting the adoption of electric appliances in NJ homes, although there is still significant potential for improvement in model accuracy. The accuracy of the trained models ranged between the range of 63-97%.

Presenting Author: Andrew Majowicz Stevens

Presenting Author Biography: Andrew is a Graduate Research Assistant at Stevens

Authors:

Andrew Majowicz Stevens
Philip Odonkor Stevens