Ai-Based Analytics and Energy Modeling Framework for Characterizing Urban Energy Systems

Abstract

Developing location-specific district energy models is essential for understanding energy patterns and providing communities with vital information to make decisions about their future energy use. However, accurately characterizing these models is challenging due to gaps in building characteristics and energy-use behavior data, as well as the labor-intensive nature of traditional modeling processes. To address this, we developed a framework that integrates top-down and bottom-up building energy data.  Using these data, the framework trains and deploys neural network models to predict missing characteristics and generate inputs for urban energy models. The framework then generates scenarios based on user-selected efficiency targets across multiple variables. Our framework integrates with URBANoptTM, a bottom-up district energy modeling platform for co-located buildings. Because data such as insulation, occupancy, and energy usage are often unavailable at the building level, we leverage census-tract-level ResStockTM data and develop methods to disaggregate it for URBANopt integration. The neural network model architecture is customized based on user-defined known and missing inputs, as well as the data modalities of these inputs. The models learn relationships between known and unknown characteristics, predicting data gaps and targeted scenarios for URBANopt models. These capabilities enable efficient urban energy characterization by generating missing inputs and creating ``What-If'' scenarios for upgrades, supporting efficiency targets. We demonstrate this methodology on a residential neighborhood in Baltimore, MD, validating it against Baltimore data in ResStock and physics-based simulations in URBANopt. Results show high predictive accuracy in data completion and scenario generation, ensuring reliability and alignment with energy dynamics. Our automated framework streamlines energy modeling and provides a reliable tool for building energy characterization.

Keywords: Urban Energy Modeling, Model Characterization, Machine Learning, Artificial Intelligence, Energy Efficiency