Neuromorphic deep learning frequency regulation in stand-alone microgrids

Abstract

Frequency instability has been a growing problem in recent years due to the rising penetration of distributed generating systems in the format of Microgrids (MGs), which are powered by renewable energy sources (RESs) with an unpredictable nature. The provision of an effective load frequency control (LFC) to an MG model under such conditions contributes significantly to the restoration regarding the unconfigured power system’s stability. An MG structure is employed, which combines renewable energy sources like solar and wind, as well as biologically renewable sustainable energy sources like wastewater, agricultural and domestic wastes, and contains appropriate battery systems and is suited to achieve both energy and waste control in this study. In addition, redox flow battery (RFB), an innovative battery storage system with a fast dynamic response in this MG structure, has been considered as one of the energy storage devices. A nonlinear integral backstepping (NIB) controller is adopted to stabilize the frequency deviation of the integrated MG system with RFB under various level of load disturbances and randomness of RESs. In particular, a spike neural network (SNN) based on neuromorphic platform is proposed to adjust the coefficients of NIB controller. Furthermore, real time analyses are performed with OPAL-RT to validate the feasibility of the proposed strategy from a systematical viewpoint. As a result of this study, the proposed controller for load disturbance, RES power changes, and contingency circumstances in bio renewable MG is determined to be flexible enough to satisfy the efficient frequency regulation for these grids.