Cyclic coordinate descent in hydrothermal nonsmooth problems (Q663492)

The paper describes an algorithm which computes the solution of a hydrothermal optimization problem involving pumped-storage plants. A hydrothermal system is composed of \(n\) thermal plants and of \(m\) hydro-plants among which the \(k\) first ones are pumped-storage ones. The optimization problem consists of minimizing the cost needed to satisfy a certain power demand during a time interval \(\left[ 0,T\right] \). The cost functional is taken as \(J(\mathbf{z})=\int_{0}^{T}L(t,\mathbf{z}(t),\dot{\mathbf{z}}(t))dt\), where \(L(t,\mathbf{z}(t),\dot{\mathbf{z}} (t))=\Psi (P_{d}(t)-H(t,\mathbf{z}(t),\dot{\mathbf{z}}(t)))\). This cost functional \(J\) has to be minimized over the set \(\Theta =\{\mathbf{z} \in (C^{1}[0,T])^{m}\mid \mathbf{z}(0)=0\), \(\mathbf{z}(T)=\mathbf{b}\), \( H_{i,\min }\leq H_{i}(t,\mathbf{z}(t),\dot{\mathbf{z}}(t))\leq H_{i,\max }\}\). Here \(z_{i}(t)\) is the \(i\)-th volume which is discharged at time \(t\) by the \(i\)-th hydro-plant and \(H_{i}\) is the effective hydraulic contribution of the \(i\)-th hydro-plant. The authors first prove an existence result for this optimization problem, under some hypotheses on the data. They then define the descent algorithm. The main result of the paper proves a global convergence result for this descent algorithm. The proof of this convergence result is deduced from a result in [\textit{W. I. Zangwill}, Nonlinear programming: a unified approach. Englewood Cliffs, NJ: Prentice-Hall Inc. (1969; Zbl 0195.20804)]. The paper ends with the description of an example where the authors use the Matematica software for the computation of the optimal solution.