One-stage R&D portfolio optimization with an application to solid oxide fuel cells

This paper provides an overview of the one-stage R&D portfolio optimization problem. It provides a novel problem model that can be solved with stochastic combinatorial optimization methods. Current solution methods are reviewed and a new method that scales to large problems, Stochastic Gradient Portfolio Optimization (SGPO), is proposed. Although SGPO is a heuristic method, we prove global convergence in certain conditions. SGPO is numerically compared to current optimization methods on a test case involving Solid Oxide Fuel Cells.