As cities strive to meet ambitious goals for reducing greenhouse gas emissions, new research is looking at how machine learning can help urban planners better understand how zoning can influence energy use. Researchers from Drexel University’s College of Engineering have developed a model that uses machine learning to make projections about the energy consumption of a given neighborhood.
The model is designed to better anticipate how energy use would change as neighborhoods evolve. This is incredibly important in Philadelphia, the oldest city in the United States, due to its widely varied building types and energy usage.
This model uses two different machine learning algorithms; Extreme Gradient Boosting (XGBoost) and Shapley Additive Explanations Analysis. XGBoost was trained on a variety of data sources including the Residential Energy Consumption Survey and the Commercial Buildings Energy Consumption Survey. The Shapley analyses allowed the team to uncover which factors had the most significant effect on energy use, for example building density, lot size and number of occupants.
The model was tested against a proposed scenario from the Delaware Valley Regional Planning Commission which activated how population increase and economic development could alter energy consumption through to the year 2045. The estimates showed residential energy use likely to decrease in lower-income areas, while mixed-income regions may see an increase in energy usage.
The researchers emphasize that their findings simply serve to provide valuable information to urban planners, rather than forming a direct link between the features and energy use changes. Their hope is that the data can be used to inform decisions about upzoning and energy reduction programs in the low-income areas where energy consumption is projected to decrease.
The model also revealed that for commercial buildings, energy use is closely linked to square footage and number of employees. Thus, commercial buildings over 10,000 square feet with a high employee count would be the primary targets of energy reduction programs.
Drexel University’s College of Engineering is leading the way in using machine learning to better understand how zoning and energy use are connected. It is hoped that this research can become an important resource for urban planners and policy makers in the future. Simi Hoque, PhD, a professor at the College of Engineering who led this research, believes this model can provide valuable advice on how zoning decisions can be used to improve energy efficiency.
