Optimization of Large Electrical Power Systems
This paper is an example of an industrial application of a well-known automatic differentiation (AD) tool for large non-linear optimizations in Power Systems. The efficiency of modern AD tools for computing first- and second-order derivatives of sparse problems, makes its use now conceivable not only for prototyping models but also for operational softwares in an industrial context. The problem described here is to compute an electrical network steady state so that physical and operating constraints are satisfied and an economic criterion optimized. This optimal power flow problem is solved with an interior point method. Necessary derivatives for the simulator of the network equations are either hand-coded or based on an AD tool, namely ADOL-C. This operator overloading tool has the advantage of offering easy-to-use drivers for the computation of sparse derivative matrices. Numerical examples of optimizations are made on large test cases coming from real-world problems. They allow an interesting comparison of performance for derivative computations.