3 présentations

• 
  15h30 - 15h55

  A tight-and-cheap conic relaxation with accuracy metrics for the ACOPF problem

  • Christian Bingane, prés., Polytechnique Montréal
  • Miguel F. Anjos, GERAD, Polytechnique Montréal
  • Sébastien Le Digabel, GERAD, Polytechnique Montréal

  Computational speed and global optimality are a key need for pratical algorithms of the OPF problem. Recently, we proposed a tight-and-cheap conic relaxation for the ACOPF problem that offers a favourable trade-off between the standard second-order cone and the standard semidefinite relaxations for large-scale meshed networks in terms of optimality gap and computation time. In this paper, we show theoretically and numerically that this relaxation can be exact and can provide a global optimal solution for the ACOPF problem.
• 
  15h55 - 16h20

  Optimal planning of long-term preventive maintenance operations on power transmission systems

  • Mariana Rocha, prés., Polytechnique Montréal
  • Miguel F. Anjos, GERAD, Polytechnique Montréal

  The interruption of service of any electrical power system equipment due to maintenance should not affect network reliability, security and continuous power supply. The selection of the optimal period to remove a transmission line from the grid temporarily and which resources (workers, vehicles, others) should be assigned to each maintenance action, constitutes the transmission maintenance scheduling problem (TMS).
  We present a mixed-integer linear program formulation of the TMS problem for a regulated electricity market, throughout a long-term period of one year and considering a yearly budget limitation. The grid is modeled as a direct current power flow network, to guarantee it meets consumers’ demand, and N-1 criteria constraints were designed to keep a reliable and safe operation.
• 
  16h20 - 16h45

  Heuristic unsupervised learning algorithm for feature engineering

  • Neda Etebarialamdari, prés.,
  • Miguel Anjos, Polytechniqe Montreal
  • Gilles Savard, Polytechniqe Montreal

  Clustering and segmentation are of great importance in many real-world tasks. In this research, we propose a heuristic algorithm in order to automatically cluster the data. The clustered labels will be added as a new feature to an industrial dataset in order to evaluate the algorithm’s performance in a forecasting task.
  Keyword: Feature engineering, heuristic, clustering