Analysis and mitigation of health impacts in generation expansion planning through linear programming and machine learning

Electrical Generation Expansion Planning (GEP) refers to the aspects of electrical grid management related to the addition, placement, transition, and intended operation of electricity generating technologies within a grid’s portfolio. The concept of GEP modeling surrounds the creation of programs and frameworks which can optimally make these decisions in a way that addresses all relevant constraints, goals, and requirements. It is clear that GEP decisions have innumerable impacts beyond simple immediate market costs. The concept of human health damages existing as an externality to GEP decision making have been almost entirely unacknowledged in the literature, outside of the work of Rodgers et al.This dissertation attempts to answer three critical questions: What are the impacts that GEP decision making has on human health? How can we incorporate these damages into our analysis and achievement of long term energy policy? What are the functional implications of human health considerate policy achievement on electrical grid functionality?These questions are answered through the development of novel frameworks and methodologies and their application to a real-life municipality. First, health impacts of various energy policies are analyzed through a bespoke GEP model and a post-hoc framework. This analysis is then extended to study the impacts that neighboring municipalities with heterogeneous energy mixtures can have on each other. The following chapter develops a model to investigate the impact that energy policy achievement has on grid functionality in the context of “day-ahead” operational planning. The model is then leveraged to further our understanding of the impacts that demand flexibility can have on the mitigation of load-shed events potentially related to heightened reliance on renewable energy, and on how we can potentially use shadow pricing as a signal. Finally, a methodology is created to endogenize health impacts directly within a GEP model to allow for active, simultaneous minimization of both incurred health damages and immediate market costs. This is achieved through the development of a novel framework allowing for the embedding of a neural-network-based surrogate model directly within a linear program.