We consider multiperiod financial risk measurement and an application to power generation optimization. In particular, we consider multiperiod extensions of the risk measure Conditional-Value-at-Risk (CVaR), which is widely used in applications due to its coherency properties. An appropriate time-consistent generalization for multiple periods is to apply CVaR-like measures recursively over the time periods. This type of risk measurement attracted interest in recent years (e.g. as nested Average VaR (nAVaR), dynamically consistent Tail VaR (DTVaR), Conditional Risk Mappings).
We discuss the case of risk measurement for stochastic value processes; risk measurement for a random variable at the time horizon can be viewed as a special case. In a finite setting in time and states on a scenario tree, we show linear formulations and the integration of the concept into multiperiod mean-risk optimization problems. As an application, we consider a power generation problem, where electricity prices are modeled by their occupation time statistics.
This is a joint work with Janos Mayer, Department of Business Administration, University of Zurich.