Dashboard for Power Outage Forecasts for an Emergency Response to Hurricanes
- Prateek Arora, Ph.D. Candidate, Data Driven Disaster Risk Analysis, NYU Disaster Risk Analysis Lab (DRAL)
- Luis Ceferino, Assistant Professor of Civil and Urban Engineering, NYU Disaster Risk Analysis Lab (DRAL)
Hurricanes damage power systems and cause power disruptions for millions of people. This project will inform utility companies and communities about the power outage risks before hurricanes makes an impact. The project’s main goal is to visualize forecasted power outages through a dashboard by developing a pipeline that utilizes National Oceanic and Atmospheric Administration weather forecasts to predict outages before a storm impacts. The outcome of this project could inform utility companies to take appropriate measures, e.g., installing backup power and placing their crews ahead of a storm to recover the power system rapidly.
Category: Urban Infrastructure
Project Description & Overview
This project will develop a visualization dashboard hosted on a webpage to project power outages for U.S. cities at risk from an incoming hurricane. The project will:
– Assess the vulnerabilities of power distribution systems to hurricanes through machine learning models that can predict outages before a hurricane makes an impact.
– Develop a visualization dashboard that can inform about forecasted power outages from an incoming storm through a visualization dashboard.
The Disaster Risk Analysis Lab (DRAL) will provide the students with the initial HTML framework of the dashboard to visualize forecasted state-wide power outages. Students will work to integrate open-source county maps to project power outages at the county level for each state. Students will develop a pipeline to:
(a) obtain the weather forecast data, e.g., wind speed and precipitation, through National Oceanic and Atmospheric Administration API. The weather forecast will be collected every three hours starting three days before a hurricane’s impact.
(b) feed the weather forecast to the machine learning-based power outage model developed at DRAL using historical data on millions of power outages.
(c) obtain the predicted outages from the machine learning model and visualize the forecasted power outages on a webpage.
The end goal is to inform the stakeholders, such as utility companies and community members, to allocate resources better and increase their resilience against prolonged outages.
Development of the power outage forecast dashboard will require the following datasets:
Historical Power Outage Data
Open Source U.S. Boundary Maps
Tigerline State Boundaries
Tigerline County Boundaries
National Oceanic and Atmospheric Administration
Knowledge of data processing in Python, including Geopandas and Excel.
Understanding of statistical and machine-learning methodologies.
Familiarity with data visualization with HTML, PHP, Ecahrt, and Leaflet or similar.
Familiarity with database tools such as mySQL.
Learning Outcomes & Deliverables
Understand the data requirements for power outage risk analysis.
Build skill sets for hands-on data visualization.
Develop risk communication and visualization skills relevant to policymaking in disaster risk management.
Dashboard visualizing the risk of power outages.
Presentation including relevant stakeholders.
Progress and final reports.