On-going Project II:
Wildfires-Energy-Resilience
[NSF CAREER: Domain-Aware Statistical Learning (CMMI 2143695)]
Topic 1: Statistical modeling for wildfire aerosols propagation using multi-source geostationary satellite remote-sensing image streams
Objectives: a physics-informed statistical model for estimating true aerosol propagation from multi-source heterogeneous remote-sensing data streams.
Application: short-term Aerosol Optical Depth (AOD) prediction for solar energy prediction (collaborating with the National Renewable Energy Lab, NREL)
Challenges: heterogeneous multi-source data with different characteristics (sampling rate, data missing rate, accuracy, etc.). The following two satellite images are taken over the same spatial area over the same time window, but clearly show different characteristics:
Aerosol Optical Depth (AOD) from NOAA Geostationary Operational Environmental Satellite (GOES) 16
Glass Fire, CA, 2020
Aerosol Optical Depth (AOD) from NOAA Geostationary Operational Environmental Satellite (GOES) 17
Glass Fire, CA, 2020
Methodology
Topic 2: Spatio-Temporal Fire Ignition Risks Modeling and Prediction on Power Grid
Objectives: to develop a statistical spatio-temporal non-homogeneous Poisson process model on linear networks, by considering dynamic environmental information (temperature, wind, humidity, ) and power grid topology.
Application: wildfire intensity modeling and prediction and risk assessment on networks power transmission lines (collaborating with the Argonne National Lab)
Illustration: The figure below illustrates the estimated (the first row) and the predicted (the second row) of the wildfire intensity on power transmission lines
