Research

Often old and outdated, frequently, urban electric grids in America are at odds with residential and commercial usage requirements. Recent weather events, such as the 2021 winter storm in Texas, have exposed serious vulnerabilities in the resilience of the electric infrastructure to support unprecedented surges in electricity demand. The growing threat of such events with climate change and the migration of populations into urban centers requires a reassessment of energy consumption patterns in most US cities. While both private providers and state regulatory bodies have detailed information on energy consumption patterns, there is a need to develop computational models to predict consumption patterns during extreme weather events, including winter storms and heatwaves. This research uses 3D models of commercial and residential buildings to create a gradient of energy usage with the granularity of a single building and propose a plan for grid distribution optimized for human-centric usage and need.

The National Oceanic and Atmospheric Administration (NOAA) has an up-to-date database of Storm Events in the US beginning January 1950. The storm events before 1995 require Optical Character Recognition to extract data from scanned documents. Some extraction has been done already, but not for every storm event. Records starting in 1996 are already in plain text. For almost all records, there are natural language narratives, and these narrative/s have the potential of containing crucial information about the episodes/events. If these narratives were in a more structured format, they could potentially yield a better understanding of destruction patterns and possible focuses for resilience. This two-part project is utilizing modern Optical Character Recognition, Information Retrieval, Deep Learning, and Natural Language Processing Tools in order to data-mine and extract information from these storm event records.

Operational Readiness for Public Health Emergencies in the U.S. (ORPHEUS) is a decision support framework that shall enable public health practitioners and responders to guide individuals, groups, communities, and populations out of the chaos brought upon a geographic region by natural or human-made disasters. Upon prioritizing regional hazards, planning authorities will determine specific mitigation strategies for those hazards. While data-driven tools that can aid regional hazard prioritization have been developed in recent years, we witness a distinct lack of such data-driven tools that facilitate the design of effective response plans capable of addressing specific hazards.

The name of this project, chimera, has two meanings, which metaphorically fit well with our project. First, a chimera is a Greek mythological, hybrid lion/goat/snake fire-breathing creature, which reflects our approach to this research. We are using an ensemble of simulation types/models to create a picture of scheduled contacts that a GSU student may make on a given day in the school semester. Another meaning for chimera is a thing that is impossible to achieve, and this is also critical to the project; No matter how close we try to simulate reality, the model will never achieve anywhere close to perfection. To combat this impossibility, we work to develop abstract methods to account for this implicit imperfection to get as close as we can without flying too close to the sun. In our research for Project Chimera, we endeavor to create an ensemble simulation model to ultimately allow us to simulate efforts to reduce the number of “scheduled” contacts a GSU student makes throughout the day.