Evaluating the utility of short-term hydrological forecasts in a hydropower system
|Authors:||Nikghalb Ashouri, Hajar|
|Abstract:||The optimal operation of a system of reservoirs is a complex decision-making problem involving, among others, the identification of a temporal trade-offs regarding the use of water. Should the last unit of water be kept in storage or rather be released for use downstream? The variability of natural inflows further complicates this decision-making problem: at any given point in space and time, this trade-off must be made without a perfect knowledge of future reservoir in flows. Generally speaking, the optimal balance between immediate and future uses of water requires the integration of short- and long-term policies. If short-term policies lead to shortsighted decisions, long-term operational strategies are not appropriate to handle short-term events such as floods. We propose a modeling framework based on the time decomposition (TD) approach: mid/long-term policies are determined first and then used as boundary conditions for the optimization of short-term policies. The mid-term optimization model captures the temporal persistence found in the weekly streamflow process whereas Ensemble Streamflow Forecasts (ESF) are used to drive the short-term model on a daily time step. More specifically, a Stochastic Dual Dynamic Programming (SDDP) generates the weekly benefit-to-go functions that are then imposed to a linear programming model implemented on each 14-days member of the ESF. This modelling framework is implemented in a rolling-horizon mode on a cascade of hydropower stations in the Gatineau River basin, Quebec, Canada. Using this modelling framework, we analyze the relationship between the economic value of different sets of short-term hydrologic forecasts. The results show that the energy generated by the hydropower system increases with the forecast's accuracy and resolution but that the relationship is not univocal; other factors seem to contribute to the forecast's utility.|
|Document Type:||Mémoire de maîtrise|
|Open Access Date:||26 January 2019|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.