Fast and efficient gradient projection algorithm for dynamic generation dispatching

A gradient projection algorithm is presented for the dynamic dispatch of thermal generations, based upon an active set strategy and exploiting second order information in determining the descent directions. The algorithm is adapted to the model of the dynamic dispatch problem which consists of the minimisation of a nonlinear objective function subject to quadratic equality constraints and linear inequality constraints. The envisaged algorithm can be applied to the dispatching of thermal units during the steep load pick-up and drop-down periods over short time intervals, and can also be used in the day-before scheduling after the unit commitment has been performed. Efficiency is demonstrated by tests on large-scale networks representative of the Italian EHV system. The limited CPU time requirements also suggest a possible on-line application in determining the constrained participation factors to be used in the Automatic Generation Control function.