Thesis Defence: Post-Disaster Emergency Electrification of Shelters and Vulnerable Households

Haofei Song, supervised by Dr. Babak Tosarkani, will defend their thesis titled “Post-Disaster Emergency Electrification of Shelters and Vulnerable Households: A Data-Driven Two-Stage Stochastic Optimization Approach” in partial fulfillment of the requirements for the degree of Master of Applied Science in Mechanical Engineering.

An abstract for Haofei Song’s thesis is included below.

Defences are open to all members of the campus community as well as the general public. Please email babak.tosarkani@ubc.ca to receive the Zoom link for this defence.

Abstract

Large-scale disasters can severely disrupt the main electricity grid, leaving shelters and vulnerable households without reliable power when electricity is most critical for safety, health, and emergency response. This thesis investigates the problem of post-disaster emergency electrification under prolonged grid outage conditions and develops an optimization-based framework for coordinating stationary and mobile energy resources. The proposed system includes shelter-level photovoltaic panels, wind turbines, battery energy storage systems, and backup generators, together with electric vehicles (EVs) that can provide mobile electricity support to shelters and critical households. The research is carried out in two stages. First, a deterministic mixed-integer linear programming model is developed to determine shelter-level resource installation decisions and post-outage operational schedules over a multi-period planning horizon. Second, this benchmark model is extended into a data-driven two-stage stochastic optimization framework that incorporates uncertainty. In particular, the stochastic framework considers uncertainty in home and shelter electricity demand, renewable resource availability, and the initial state of charge of EV batteries. In the stochastic model, representative demand scenarios and associated probabilities are generated for homes and shelters using a data-driven procedure based on Gaussian mixture models, principal component analysis, robust kernel density estimation, uncertainty filtering, and scenario construction. In addition, renewable resource availability is represented through forecasting-based profiles for wind speed and solar irradiance, and uncertainty in EV initial state of charge is incorporated as a scenario-dependent parameter. The proposed models are evaluated through a case study motivated by the impacts of Typhoon Faxai in Chiba Prefecture, Japan. The results show that coordinated use of distributed generation, battery storage, backup generation, and EV-based mobile electricity support can improve service continuity, reduce unmet demand, and lower emergency operating costs. The findings also highlight the value of EVs as flexible supplementary energy resources in disaster settings and demonstrate the importance of incorporating uncertainty into emergency energy planning. This thesis contributes an integrated decision-support framework that links disaster-resilient microgrid planning, EV-enabled mobile energy support, and uncertainty modelling and stochastic optimization for post-disaster emergency electrification.