Bundle methods applied to the unit-commitment problem

On a daily-time scale, the power generation management poses a large-size and composite problem. For this reason, price decomposition has become one of the most commonly used strategies in this field. Classical spatial decomposition, obtained by dualizing the coupling constraints, is not the only way of decomposition. Some recent tests have shown that a space/time decomposition is particularly well-suited to deal with transmission constraints. The resulting dual problem is large but tractable (more than ten thousand variables), while the spatial decomposition would yield exceedingly many dual variables. This paper is devoted to the application of bundle methods to such large-scale nondifferentiable problems. To assess the approach we present some numerical tests which show that, in practice, bundle methods can cope with such large problems.