Diagnosing the Time-Varying Value of Forecasts in Multiobjective Reservoir Control

Abstract

This work forwards time-varying sensitivity analysis of control policies as an approach to identify the value of the alternative timing of forecasts in a complex multipurpose reservoir system. The time-varying forecast diagnostics are demonstrated using the Conowingo Dam in the Lower Susquehanna River Basin. First, we evaluate the utility of multiple lead time forecasts, both in isolation and in combination, to determine the sensitivity of optimal operations to forecast lead time. We then use time-varying sensitivities to determine how the system utilizes a particular forecast through time for two policies that emphasize hydropower and environmental objectives, respectively, in order to determine how the value of inflow information changes with respect to stakeholder preferences and evolving hydrologic conditions. Results show that the value of inflow information varies greatly depending on the objectives being emphasized and can decline or fade entirely when system constraints (e.g., lower bounds on reservoir stage and spillway thresholds) inhibit the use of forecasts to inform decisions, particularly during extreme events.