Decision Support System for the hydropower plants management: the MINERVE project

The MINERVE project was initiated for flood management in the Upper Rhone river basin. Its first goal is the prediction of floods. The second one is the optimum management of the complex system of hydropower plants located in the basin in order to diminish the peak discharge in the Rhone River and its tributaries. The MINERVE Interactive Decision Support (MINDS) proposes turbine and bottom outlet preventive operations before the flood peak, for having storage volume in the reservoirs during the peak flow. The iterative ranking Greedy algorithm allows successive resolutions for all the hydropower plants in an iterative way until reaching the optimum operations of the whole complex. The multi-criteria analysis optimises the start and the end of the preventive operations (turbine and bottom outlet operations) for the whole ensemble hydrological forecasts for a selected hydropower plant. Dams and Reservoirs under Changing Challenges – Schleiss & Boes (Eds) © 2011 Taylor & Francis Group, London, ISBN 978-0-415-68267-1 2 HYDROLOGICAL FORECASTS 2.1 Meteorological input data The MINERVE system exploits probabilistic COSMO-LEPS weather forecast as well as deterministic weather forecasts COSMO-7 and COSMO-2, all of them operated by MeteoSwiss. COSMO-LEPS is the limited-area EPS (Ensemble Prediction System) developed within the COSMO consortium (Consortium for Small-scale Modeling) and combines the benefits of the probabilistic approach with the high-resolution detail of the model. Twice per day, COSMOLEPS provides high resolution probabilistic forecasts (horizontal mesh-size of 10 km) based on a 16-member ensemble for central and southern Europe with a lead time of 120 h. Deterministic forecasts COSMO-7 and COSMO-2 are a support of COSMO-LEPS. The regional COSMO-7 is driven by the global model of ECMWF (European Centre for MediumRange Weather Forecasts) and covers most of western and central Europe. It is computed on a grid spacing of about 6.6 km and is calculated twice per day for 72 h lead time. The local COSMO-2, driven by COSMO-7, covers the Alpine region with Switzerland located at the center and is computed on a grid spacing of about 2.2 km. It is calculated 8 times a day for a 24 h lead time. Both of them allow the use of the most recent atmospheric conditions observed and the benefit of the short range forecasting. 2.2 Hydrological model The catchment area of the Rhone River has about 5500 km and has been divided in 239 subcatchments taking into account all hydraulic structures of dams and hydropower plants. The hydrological concept used to estimate the discharge in the outlet of each sub-catchment is based on the GSM-Socont (Glacier & SnowMelt -SOil CONTribution) model (García H. et al., 2007). For the hydrological part of modelling, each sub-catchment is divided into two parts, glacier and non-glacier, both divided in altitude bands. Precipitations and temperatures values are calculated from the meteorological input data for each one of these altitude bands. Then, a snow model follows the temporal evolution of the height and saturation degree of the snow. The snow melt produces an equivalent precipitation starting from a rate of saturation threshold. In the case of a non-glacier band, this equivalent precipitation supplies the infiltration and the transfer model, composed by two parallel non-linear reservoirs, which produce the slow and fast components of the discharge going to the outlet of the sub-catchment. In the case of a glacier band, the equivalent precipitation resulting from the snow melt is transferred to the outlet by a linear reservoir. When there is no more snow, a glacier model produces (when temperature is higher than zero) a discharge which is also transferred to the outlet of the sub-catchment by a linear reservoir behaviour. 2.3 Hydrological simulations This semi-distributed hydrological model was built using the hydrological and hydraulic simulation tool Routing System II (García H. et al., 2007; Jordan et al., 2008). This software was designed to simulate the formation and the propagation of free surface flows in a complex system. It allows hydrological and hydraulic modelling by an oriented object approach, according to a semi-distributed conceptual scheme. It takes into account special hydrological processes such as snow and glacier melt, surface and sub-surface flows, routing in reservoirs, water transfer tunnels and rivers as well as the modelling hydraulic structures with valves, gates, water intakes, turbines or pumps. An evolution of Routing System II tool provides a flood forecast in real-time, coupling the observed measurements and the weather forecast information with the hydrological model. Every time a new weather forecast is provided, the hydrological forecast is updated. Dams and Reservoirs under Changing Challenges – Schleiss & Boes (Eds) © 2011 Taylor & Francis Group, London, ISBN 978-0-415-68267-1 After the first developments of the project with deterministic forecasts until 2006 (Jordan, 2007), the MINERVE system is enhanced since 2008 by implementing new probabilistic forecasts (García H. et al., 2009a) as well as other improvements in other domains (García H. et al., 2009b) with the aim of providing better hydrological forecasts to the decision support system. 3 MINERVE INTERACTIVE DECISION SUPPORT 3.1 Hydraulic model The hydraulic model of the Wallis and Vaud Cantons, developed for the optimisation tool MINDS (MINERVE Interactive Decision Support), is a simplified model of this complex river basin (Fig. 1). It contains the most important reservoirs RES (triangles), with its bottom outlets and spillways (square dotted lines), hydropower plants HPP (round doted lines), as well as the main river network (solid lines) with the main control points CP (big circles). Mattmark Grande Dixence Gebidem Ferden Zeuzier Mauvoisin Emosson Salanfe Toules Moiry Turtmann SCEX OFEV