Modeling and simulation of a microgrid using feedforward neural networks

Abstract

Electric power grids and complex computer systems have many similar properties of the operation behavior and the structure. A microgrid can be treated as a small electric grid that contains consisted of numerous residential loads, energy storage units, and distributed energy. The goal of implementing microgrids is to supply power to homes even in the event of an electric grid outage. That is, the stored energy in the storage unit and distributed generation will supply energy to the load until the main grid return to the normal operation, and therefore, supply power to the load and store energy back to the storage unit. This method allows decentralization of the electric grid regarding control and energy supply. To deal with decentralized systems, one needs to construe the electric grid as a system of systems (SoS), and use models that can capture the dynamics of the microgrid. This paper presents a model of microgrid using feedforward neural networks. This model can be utilized in complex system modeling techniques such as agent-based approaches and system dynamics, or a combination of various methods to represent different electric elements. An example of modeling real microgrid is presented to demonstrate the emergent characteristics of the interconnected system. Simulation results and waveforms are discussed.