Planning & Management
394638 - Direct policy search for multiobjective financial risk management in snow dominated hydropower systems
Wednesday, June 6
8:30 AM - 10:00 AM
Location: Northstar B
(1)Department of Environmental Sciences and Engineering and (2)Center on Financial Risk in Environmental Systems, Gillings School of Global Public Health and UNC Institute for the Environment, University of North Carolina at Chapel Hill
Gregory Characklis, Chapel Hill, NC, USA – University of North Carolina at Chapel HIll; Patrick Reed, Ithaca, NY, USA – Cornell University
Hydropower generators in snow dominated systems are vulnerable to generation shortfalls during drought, giving rise to financial risk. Additionally, like all power generators, they must consider the effects of temperature and natural gas price on electricity demand and wholesale prices. Generators typically have large fixed operating costs and high levels of debt service, so low revenue periods associated with these uncertain conditions can be highly disruptive. Using the Hetch Hetchy Power system, owned and operated by the San Francisco Public Utilities Commission, as a case study, it has been shown that an index insurance contract based on snow water equivalent depth can be used to effectively hedge some of the supply risk associated with low snowfall. In this research, a comprehensive portfolio of financial risk management tools is developed in order to hedge the collective and related financial effects of variable snowpack, temperature and natural gas price. The portfolio is composed of snowpack-based index insurance contracts, as well as a reserve fund (i.e. self-insurance) and financial derivatives based on natural gas price and temperature. The strategy for assembling a portfolio each year is optimized using an Evolutionary Multiobjective Direct Policy Search. This search method is a closed-loop control strategy in which decision outputs (e.g. How much insurance to buy this year?) are parameterized as functions of state variables and stochastic inputs. Tradeoffs in objectives such as expected revenues, minimum revenues, revenue volatility, and portfolio complexity are considered by generating a Pareto-approximate set of policies using the Borg Multiobjective Evolutionary Algorithm.