Acquiring Constraints for a Non-linear Transmission Maintenance Scheduling Problem

Hugo Barral, Mohamed Gaha, Amira Dems, Alain Côté, Franklin Nguewouo, Quentin Cappart

Abstract: Over time, power network equipment can face defects and must be maintained to ensure transmission network reliability. Once a piece of equipment is scheduled to be withdrawn from the network, it becomes unavailable and can lead to power outages when other adjacent equipment fails. This problem is commonly referred to as a transmission maintenance scheduling (TMS) problem and remains a challenge for power utilities. Numerous combinatorial constraints must be satisfied to ensure the stability and reliability of the transmission network. While most of these constraints can be naturally formalized in constraint programming (CP), there are some complex constraints like transit-power limits that are challenging to model because of their continuous and non-linear nature. This paper proposes a methodology based on active constraint acquisition to automatically approximate these constraints. The acquisition is carried out using a simulator developed by Hydro-Québec (HQ), a power utility to compute the power-flow of its transmission network. The acquired constraints are then integrated into a CP model to solve the HQ network’s TMS problem. Our experimental results show the relevance of the methodology to approximate transit-power constraints in an automated way. It allows HQ to automatically schedule a maintenance plan for an instance that remained intractable until now. To our knowledge, it is the first time that active constraint acquisition has been used successfully for the TMS problem in an industrial setting.