Daily Generation Scheduling : Decomposition Methods to Solve the Hydraulic Problems

TL;DR

This paper introduces two novel decomposition methods for solving complex short-term hydro-generation scheduling problems, improving feasibility and economic efficiency in hydroelectric power management.

Contribution

It proposes two new decomposition algorithms, one based on prediction and primal-dual relaxation, and another on price decomposition with Augmented Lagrangian, applied to hydraulic scheduling.

Findings

Algorithms demonstrate high efficiency in numerical tests.

Methods improve the process at Electricité de France.

Solutions are feasible and economically satisfying.

Abstract

Short-term hydro-generation management poses a non-convex or even non-continuous optimization problem. For this reason, the problem of systematically obtaining feasible and economically satisfying solutions has not yet been completely solved. Two decomposition methods, which, as far as we know, have not been applied in this field, are here proposed : $\bullet$ the first is based on a decomposition by prediction method and the coordination is a primal-dual relaxation algorithm, $\bullet$ handling the dynamic constraints by duality, the second achieves a price decomposition by an Augmented Lagrangian technique. Numerical tests show the efficiency of these algorithms. They will enable the process in use at Electricit{\'e} de France to be improved.