Research Gateway

Abstract

grid power feeding is very difficult and uneconomic at the remotely located loads like hilly area and Island. For such locations, distributed power generation is practical option. The wind energy conversion system (WECS) can be considered as distributed power generation in standalone application. However the major problems associated with WECS are inconsistent power quality and unreliable power generation in changing environment. This paper proposed special form of recurrent neural network type echo state network based control algorithm which is used to mitigate various current, voltage and frequency related power quality problems. This type of network requires only one layer training i.e. the output weight matrix update is required during the entire training process. It is used to extract fundamental active as well as reactive weight component from distorted load currents. Further, these components are used to estimate the reference source current such that generator only supply real power. It is followed by generation of gate pulses for the Distribution Static Compensator (DSTATCOM) which is used as an active shunt filter in standalone WECS to maintain the power quality as per IEEE-519-2014 standard. The entire setup of SWECS is developed in MATLAB environment followed by experimental validation of control algorithm.