Dynamic energy management in smart grid: A fast randomized first-order optimization algorithm

A crucial issue in the smart grid is how to manage the controllable load resources of end-users, in order to reduce the economic costs of system operation and facilitate to utilize renewable energies. This paper investigates a fast randomized first-order optimization method to explore the solution of dynamic energy management (DEM) for the smart grid integrated large-scale distributed energy resources. A complicated time-coupling and multi-variable optimal problem is presented to determine the load scheduling for the electricity customers. The main challenge of the proposed problem is to enable the efficient processing of the large data volumes and optimization of aggregated data involved in DEM. The first-order method as one of big data optimization algorithms is able to exhibit significant performance for computing globally optimal solutions based on randomization techniques. Using such solution approach, we can reformulate the original problem into an unconstrained augmented Lagrangian function. The optimal results can be obtained from computing the gradient based on the information of the first-order derivative. To speed up the calculations of obtaining the feasible solutions, the optimization variable matrix used to update the Lagrangian multiplier can be replaced with the corresponding low-rank representation in the iteration process. Both theoretical analysis and simulation results suggest that the proposed approach may effectively solve the optimal scheduling problem of DEM considering users' participation.