Harmonic Mitigation of the Microgrid Using Multi Hidden Layer Recurrent Neural Network Controlled SAPF

Abstract Recently, renewable energy systems are more prevalent than traditional energy systems. Especially, Photovoltaic (PV) systems and wind power systems are playing an important part in meeting the world's energy needs. The present needs of more nonlinear loads drastically create power quality issues in the grid-connected system. As a result, it causes an undesired power quality defect that takes the manner of an alteration in the amplitude and patterns of voltage and current in the power system. The shunt active power filter (SAPF) delivers the appropriate current in shunt to grids while suppressing harmonics created by grid-tied nonlinear loads. This paper proposes a multi hidden layer recurrent neural network (RNN) controller for the developing a reference signal in synchronous reference frame theory. RNN is proposed for its high efficacy compared to neural networks, due to self-loops and memory. The effectiveness of the proposed multi-hidden layer RNN-based SAPF system is compared with the performance of neural networks and traditional PI controller-based SAPF systems in terms of harmonic mitigation. Effective harmonic mitigation by proposed multi-hidden layer RNN results current THD of proposed microgrid to 1.74%.