H31J-0769:
Incorporating Hydrologic Routing into Reservoir Optimization Model: Implications for Hydropower Production Planning

Wednesday, 17 December 2014
Nicholas Zmijewski1, Anders L E Worman1 and Andrea Bottacin-Busolin2, (1)KTH Royal Institute of Technology, Stockholm, Sweden, (2)University of Manchester, School of Mechanical, Aerospace and Civil Engineering, Manchester, United Kingdom
Abstract:
Renewable and intermittent energy sources are likely to become more important in the future and consequently lead to a change in the production strategies of hydropower to account for expected production fluctuations. Optimization models are used as a tool to achieve the best overall result in a network of reservoirs and hydropower plants. The computational demand increases for large networks, making simplification of the physical description of the stream flows necessary. In management optimization models, the flows behavior is often described using a constant time-lag for water on flow stretches, i.e., the release of water mass from an upstream reservoir is time-shifted as inflow to the subsequent reservoir. We developed an optimization model that included the kinematic-diffusion wave equation for flow on stretches, which was used to evaluate the role of the model improvement for short term production planning in Dalälven River, a study case with 36 hydropower stations and 13 major reservoirs . The increased complexity of the time-lag distributions of the streams resulting from the kinematic-diffusion wave equation compared to the constant time-lag model was found to be highly important for many situation of hydropower production planning in a regulated water system. The deviation of optimized power production resulting from the two models (time-lag and kinematic-diffusive) increases with decreasing Peclet number for the flow stretches – the latter being evaluated for all included stretches. A procedure emulating the data-assimilation procedure present in modern systems, using the receding horizon approach, was used in order to describe the dynamic effect of the resulting flow prediction deviation. The procedure also demonstrated the importance of high frequency data assimilation for highly effected streams, which implies that the error in predicted power production decreases with decreasing time step of forecast updating.