4 Presentations

• 
  03:30 PM - 03:55 PM

  A successive linear programming algorithm for the stochastic reservoir management problem

  • Charles Gauvin, presenter, Polytechnique Montréal
  • Erick Delage, GERAD, HEC Montréal
  • Michel Gendreau, Polytechnique Montréal

  This talk considers optimal electricity generation for a multireservoir hydroelectrical complex in the presence of uncertain inflows. We consider affine decision rules and propose a successive linear programming heuristic to obtain good solutions rapidly. We embed this approach into an existing simulation and rolling optimization framework with linear time series.

• 
  03:55 PM - 04:20 PM

  Coordinated hydropower maintenance & operation scheduling via decomposition methods

  • Jesus Rodriguez, presenter, HEC Montreal
  • Pascal Côté, Rio Tinto
  • Miguel F. Anjos, GERAD, Polytechnique Montréal
  • Charles Audet, GERAD - Polytechnique Montréal

  Generator maintenance scheduling is a difficult combinatorial problem with significant impact on the system operation. Since the simultaneous solution of the hydro maintenance and operation is hard to achieve, we present a decomposition approach for this problem. Our model formulation is based on the real case of Rio Tinto Aluminium.

• 
  04:20 PM - 04:45 PM

  Updating reservoir management policy using hydrological modeling and stochastic dynamic programming

  • Alexandre Martin, presenter, Université de Sherbrooke
  • Pascal Côté, Rio Tinto
  • Robert Leconte, Université de Sherbrooke

  We present a case study of the Nechako Reservoir, B. C., Canada, on which management rules are updated using on-line SDP and streamflow forecasts produced with a distributed hydrological model. When compared with a classical off-line SDP policy, results show a reduction in unproductive spills and downstream flood events.

• 
  04:45 PM - 05:10 PM

  Dynamic stochastic models with monotone optimal policies for one, two or many reservoirs

  • Bernard Lamond, presenter, Université Laval

  We review several monotonocity properties of simplified optimization models for hydroelectric systems with one, two or more reservoirs where the rewards are concave functions of electricity production and the natural inflows are stochastic. We examine various situations such as one or many reservoirs, head variation or not, same or different routings for spilled and turbined flows, and time sequence of decisions.