Reinforcement learning control approach for autonomous microgrids. (2nd January 2021)

• Record Type: Journal Article
• Authors: Mahmoud, M. S.; Abouheaf, M.; Sharaf, A.
• Is Part Of: International journal of modelling & simulation. Volume 41:Number 1(2021)
• Page Start: 1
• Page End: 10
• Publication Date: 2021-01-02
• Subjects: Renewable energy -- microgrids -- reinforcement learning -- adaptive critics -- neural networks -- optimal control; Mathematical models -- Periodicals; Simulation methods -- Periodicals; Mathematical models; Simulation methods; Periodicals; 003.3
• DOI: 10.1080/02286203.2019.1655701
• Languages: English
• ISSNs: 0228-6203

Abstract

ABSTRACT: The increasing penetration of the renewable energy systems into the main power grids has raised concerns about robustness of the existing control mechanisms. An adaptive learning approach is proposed to regulate the output voltage of an autonomous distributed generation source. This controller solves the optimal control problem for that generation source by finding a recursive solution for the underlying Bellman optimality equation. A value iteration algorithm is introduced in order to find the optimal control strategy in a dynamic learning environment. Means of adaptive critics are employed to implement the solution without knowing the drift dynamics of the microgrid. The developed controller is shown to be robust against different power system disturbances and exhibited competitive behavior when compared to a standard Riccati control approach subject to uncertain dynamical environment.