joint work with Napat Rujeerapaiboon, Daniel Kuhn, Wolfram Wiesemann
Peak/off-peak spreads in European electricity spot markets are eroding due to the nuclear phaseout and the recent growth in photovoltaic capacity. The reduced profitability of peak/off-peak arbitrage forces hydropower producers to participate in the reserve markets. We propose a two-layer stochastic programming framework for the optimal operation of a multi-reservoir hydropower plant which sells energy on both the spot and the reserve markets. The backbone of this approach is a combination of decomposition and decision rule techniques. Numerical experiments demonstrate its effectiveness.
joint work with Napat Rujeerapaiboon, Daniel Kuhn
The revenue-maximizing mechanism for selling I items to a buyer with additive and independent values is unknown, but the better of separate item pricing and grand-bundle pricing extracts a constant fraction of the optimal revenue. We study an online version of this pricing problem, where the items are sold to (each of) T identical and independent buyers arriving sequentially. We propose a two-layer multi-armed bandit algorithm that is reminiscent of the UCB algorithm and offers an asymptotic constant-factor approximation guarantee relative to the (unknown) best offline mechanism.
joint work with Peyman Mohajerin Esfahani, Viet Anh Nguyen, Soroosh Shafieezadeh Abadeh
Existing tractability results for distributionally robust optimization problems with Wasserstein ambiguity sets require the nominal distribution to be discrete. In this talk we derive tractable conservative approximations for problems with continuous nominal distributions by embedding Wasserstein ambiguity sets into larger Chebyshev ambiguity sets. We identify interesting situations in which these approximations are exact.