Grid-constrained optimal predictive power dispatch in large multi-level power systems with renewable energy sources, and storage devices

This paper presents a novel approach for the predictive power dispatch of a large number of power system units that can be diverse in their power as well as power ramp ratings and can be dispersed both in distribution grids and transmission networks. In particular, the involved unit portfolio consists of Renewable Energy sources (RES), conventional generation sources, flexible loads, and storage devices, which are modeled using the Power Nodes modeling framework. Unlike aggregation methods which do not consider the position of a power system unit in the grid topology, our centralized optimal power dispatch strategy with a combined AC Optimal Power Flow (AC-OPF) explicitly accounts for grid constraints. The multi-period dispatch problem induced by the energy storage devices is solved by a predictive power dispatch scheme based on Model Predictive Control (MPC). We further propose a distributed three-stage optimization process that hierarchically divides the original dispatch problem into independent sub-problems according to the grid structure which are solvable in parallel. As a result, we obtain a significant complexity reduction with respect to the original centralized dispatch optimization problem, such that we can calculate the power dispatch of a large number of units with reasonable computational effort. Finally, based on benchmark grids of different sizes, we demonstrate the improved performance of the here proposed distributed approach over the original centralized dispatch optimization approach and show that the simulation time is reduced for large systems as compared with the centralized approach.