Hydropower operations optimization models select a sequence of releases from one or more reservoirs that maximizes the expected benefit while honoring many social and environmental constraints. A 3-tiered time-decomposition algorithm is adopted to compute optimal sub-daily releases for the Harris Station reservoir in Maine, USA. This involves solving nested optimization models, each with a different planning horizon and time-step, where the longer-term planning models inform the shorter-term models. This allows for rapid optimization of short-term operations, while efficiently considering seasonal objectives and constraints. In the case study presented, 6-hr release decisions in a weekly model are made by iteratively solving weekly, monthly, and annual models using sampling stochastic dynamic programming. A key consideration is how uncertainty is represented in each of the nested models. Uncertainty is inherent in hydropower operations optimization because the future availability of water and future energy prices are unknown at the time a decision is made. In order to ensure efficient operation of the hydropower system, it is often important that such uncertainties be well represented. Reservoir operations are simulated using release decisions from time-decomposition models with different representations of uncertainty. By comparing the operational efficiency of each model, the relative merits of different uncertainty representations are examined. In particular, we consider the value of inflow forecasts to inform the uncertainty model at various planning horizons and how this changes with seasonal hydrology. Summer inflows are generally low, and it is often desirable to operate at full head to maximize generated power per volume released. Still, brief and intense localized rainstorms can cause a spike in reservoir inflow, which can result in spilling. Not surprisingly we found that short-term forecasts are of most importance to summer reservoir performance, and longer-term forecasts contributed little to operational efficiency. In other periods of the year the relative importance of long- and short-term forecasts varies.
Lamontagne, Jonathan R. and Stedinger, Jery Russell, "Short-Term Hydropower Optimization Using A Time-Decomposition Algorithm" (2014). CUNY Academic Works.