Different time-scale approaches to the real power dispatch of thermal units

Different time-scale approaches to the dispatch problem are proposed for the centralized control of real power. In particular, the problem of optimally sharing a time-varying load among the committed thermal units of an electrical power system is addressed. It is shown that the real power dispatch of the thermal units is handled at different time scales as a dynamic problem where rate of change limits , environmental constraints, and the security constraints of the traditional static approach are taken into account. The adoption of the same general model guarantees the coherency of the results in the three phases of day-before scheduling, advance dispatch, and online dispatch. A modified gradient projection algorithm, which suitably exploits some good second-order information, has been adopted for the solution of the large-scale problem arising from the discretization of the dynamic model. The resulting procedure is capable of solving the very large day-before scheduling (more than 2000 variables and 4000 constraints for the Italian EHV system) in a few minutes, as well as providing the solution of the advanced dispatch problem in no more than 10 s even when a corrective rescheduling is required