Search

Multimethod Optimal Power Flows have been developed for future real time use in the French national energy control center. They must be robust, accurate and fast, with a statement more intricate than usual, e.g. including contingency analysis and constraints. In order to meet these requirements, the basic frame of the robust and accurate Differential Injections method is followed, and complementary methods are combined with this frame in order to speed up algorithms. Basically, non-linear active and reactive reduced models, expressed in terms of control variables only, are iteratively built up then optimized by the Generalized Reduced Gradient process. But other algorithms are combined with this basic frame, such as Constrained Reactive Implicit Coupling (CRIC), quadratic programming, “implicit Hessian” technique introducing new sparsity techniques into quadratic developments, and a sparse reduced gradient method improving initial voltage magnitudes. “The right algorithm in the right place” was the rule applied in this work. Numerical results correspond to the present 650 node French network, for which 3 seconds only are necessary on a 25 MIPS computer to perform an usual active reactive OPF. With respect to the basic frame, speed has been increased by a factor 10, providing convenient tools for real time use.

Multimethod Optimal Power Flows have been developed for future real time use in the French national energy control center. They must be robust, accurate and fast, with a statement more intricate than usual, e.g. including contingency analysis and constraints. In order to meet these requirements, the basic frame of the robust and accurate Differential Injections method is followed, and complementary methods are combined with this-frame in order to speed up algorithms. Basically, non-linear active and reactive reduced models, expressed in terms of control variables only, are iteratively built up then optimized by the Generalized Reduced Gradient process. But other algorithms are combined with this basic frame, such as Constrained Reactive Implicit Coupling (CRIC), quadratic programming, "implicit Hessian" technique introducing new sparsity techniques into quadratic developments, and a sparse reduced gradient method improving initial voltage magnitudes. "The right algorithm in the right place" was the rule applied in this work. Numerical results correspond to the present 650 node French network, for which 3 seconds only are necessary on a 25 MIPS computer to perform an usual active reactive OPF. With respect to the basic frame, speed has been increased by a factor 10, providing convenient tools for real time use.