Stochastic Multizonal Electricity Generation Capacity Expansion Planning Under Weather and Demand Uncertainties

Abstract

Investment decision-making in the electricity sector is a complex process due to the inherent long-term uncertainties found at every level: weather conditions, fuel prices, demand growth, and technological advancements. Generation Expansion Planning (GEP) models determine the technology of power generation units to be constructed, their capacity, and the year of construction to ensure that projected demand is met throughout the planning horizon. Open research questions in generation expansion planning include the choice of the size of the representative periods used to reduce model complexity and the way multi-year historical weather and demand data are incorporated into the analysis. Stochastic optimization is a branch of optimization that incorporates uncertainties within an optimization model by defining uncertain scenarios with probabilities of occurrence. In this thesis, we develop a stochastic multizonal electricity generation expansion planning model focused on addressing weather- and demand-related uncertainties. The analysis compares different sets of input scenarios that vary by the size of representative periods or incorporate multiple historical meteorological years against deterministic approaches. We show that stochastic optimization leads to minimizing the risk of choosing biased sample data in terms of net present costs and unmet demand compared to deterministic approaches based on single data sets or employing scenario averaging techniques. Other issues related to the effects of incorporating operational details on investment decisions are further discussed and analyzed along with their practical implications.