Rishee Jain, Alex Nutkiewicz, Zheng Yang
Urban building energy models are emerging tools meant to analyze and understand the energy performance of multiple buildings within a dense urban area. However, accurate performance prediction of these models remains a challenge because of their inability to account for the intra-building energy dynamics and interdepenencies that can have a substantial impact on building energy use.
The advent of smart meters and open data initiatives has resulted in an explosion of structured and unstructured data streams describing buildings in their urban environment. DUE-S is a framework for integrating network-based machine learning algorithms into a traditional energy simulation workflow in order to better understand the energy consumption of buildings in urban environments. A hybrid model of this kind will enable more accurate energy performance characterization at multiple temporal and spatial scales.
The world is rapidly urbanizing. And as more people move to live and work in cities, they are becoming increasingly responsible for the world's primary energy usage and greenhouse gas emissions, with the built environment being the largest contributor to these issues. Urban buildings represent a tremendous opportunity to enhance the sustainability of cities as 90% of urban buildings are estimated to be energy inefficient and up to 30% of a building's energy consumption is wasted .
One of the key challenges to enhancing the energy efficiency of urban buildings is the lack of accurate energy performance prediction models. These models fail to capture a building's urban context, and without accurate performance characterization and prediction, designers and engineers struggle to assess the energy, environmental, and economic implications of their early-stage design and retrofit decisions thus failing to shape a building's energy usage for its entire lifecycle. This challenge is further exacerbated as adjacent buildings and the overall urban area become increasingly energy intensive, resulting in substantial energy, environmental, and monetary impacts on not only the modeled building, but the ones around it as well.
During a building's pre-design phase, its energy performance is often evaluated using an energy simulation that does not consider the effects of its surroundings. However, as is the case for many buildings in dense cities, they can be signifcantly impacted by the diverse and convoluted urban context due to effects such as shading, thermal transfer, and balance, fluid dynamics, urban heat island effect, and district level service. By not accounting for these impacts of urban context and inaccurately estimating its performance by even 10%, a large, mixed-use office building may end up consuming an extra 2.8M kWh in energy , producing an additional 1.5M lbs of CO2 emissions . and spending $670k in added costs each year , as well as having additional second-order impacts on its surrounding buildings.
The proposed research consists of the following objectives:
1. Construct a simulation-based urban energy model for a dense test bed of buildings
2. Train and evaluate the prediction accuracy of a network-based machine learning algorithm on the test bed of buildings
3. Validation of DUE-S workflow in collaboration with CIFE industry collaborator
The most significant outcome of this project is a Data-driven Urban Energy Simulation (DUE-S) that leverages the computational intelligence of machine learning algorithms and the interpretability of physics-based energy simulation. We expect this model and its associated workflow will be generalizable for any city or dense building portfolio and will be able to simulate energy consumption on an individual, community, and urban scale.
By visualizing the energy usage of buildings across a city, policymakers will have better awareness of the effects of new citywide interventions. Designers and building operators will understand the effects of energy consumption and indoor environmental quality on not only their building, but surrounding ones as well. And as data is being increasingly used in the planning of newer, smarter cities, DUE-S will be able to employ urban data streams from existing cities to optimize building energy use and to ehlp inform key decisions related to energy efficiency in both early stage design and retrofit programs throughout a building's lifecycle.
Over the past year, the research team has successfully completed one case study on 22 buildings on a university campus in California in order to demonstrate the feasibility of using this type of integrated model to characterize the energy performance of buildings at multiple temporal and spatial scales. The research team has had its preliminary results presented at the 2017 International Conference on Applied Energy  and subsequently published as a full-length journal paper in Applied Energy . Our next steps include conducting additional case studies on more additional urban morphologies to further validate our modeling approach and developing a method to implement DUE-S in a real-world building and urban scale retrofit analysis to support sustainable urban planning and operations.
Final Project Report
TR228 - Data-driven Urban Energy Simulation (DUE-S): A framework for integrating engineering simulation and machine learning methods in a multi-scale urban energy modeling workflow, Nutkiewicz, Yang, Jain
 Breaking Energy (2011). “The Top Ten Ways We Waste Energy and Water in Buildings.” <http://breakingenergy.com/2011/07/26/the-top-ten-ways-we-waste-energy-and-water-in- buildings/>
 San Francisco Dept. of Environment (2014). “San Francisco Existing Commercial Buildings Performance Report.” <https://sfenvironment.org/sites/default/files/fliers/files/sfe_gb_ecb_performancereport.pdf>
 PG&E. “Carbon Footprint Calculator Assumptions.”<https://pge.com/includes/docs/pdfs/about/environment/calculator/assumptions.pdf>
 PG&E. “Tariffs.” <https://www.pge.com/tariffs/electric.shtml#COMMERCIAL>
 A. Nutkiewicz, Z. Yang, and R. K. Jain, “Data-driven Urban Energy Simulation (DUE-S): Integrating machine learning into an urban building energy simulation workflow,” in Energy Procedia, 2017, vol. 142, pp. 2114–2119.
 A. Nutkiewicz, Z. Yang, and R. K. Jain, “Data-driven Urban Energy Simulation (DUE-S): A framework for integrating engineering simulation and machine learning methods in a multi-scale urban energy modeling workflow,” Applied Energy, vol. 225, pp. 1176–1189, Sep. 2018.