UBC Theses and Dissertations
Parameter-space dependence in a stochastic dual dynamic programming model of a multireservoir hydropower system Mitchell, Kyle Terrence
Stochastic Dual Dynamic Programming is a heuristic stochastic optimization algorithm used to analyze multistage decision making problems with stochastic variables through Benders decomposition and solvable linear approximations. Benders cuts are utilized to develop a piecewise linear approximation of a 'benefit-to-go' or future value function at each stage. This thesis assesses BC Hydro's Columbia and Peace River hydroelectric network representation in a multistage stochastic optimization problem with a finite planning horizon. The model attempts to maximize the 'water value' across the MCA, GMS, and ARD facilities over a duration of five years. Existing model parameter-space is heavily restricted due to the use of static market prices and load demands. Changes have been proposed to expand on the current parameter-space pertaining to inflow dependent market electricity prices through the implementation of seasonally forecasted price and load blocks. Preliminary results indicate the existing model implementation does not provide sufficient parameter-space coverage with respect to market prices, resulting in possible over-estimations of water value as the model approaches convergence.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International