Variable learning adaptive gradient based control algorithm for voltage source converter in distributed generation

This study presents an adaptive control algorithm known as variable learning and adaptive gradient based least mean square for improving the power quality features in standalone distributed generation. Further, frequency and voltage are regulated to set reference value at the terminal of the self-excited single-phase induction generator running in isolated mode. The variable learning and gradient-based least mean square (VLGLMS) algorithm is insensitive to its gradient, step size and sensors noise unlike least mean square algorithm whose convergence performance is influenced by the step-size parameter. In this study, VLGLMS is utilised to compute the active and reactive weights of fundamental load current for estimation of the reference source current. The sinusoidal reference current estimation is followed by generation of gate pulses for operating the DSTATCOM for improving the power quality features of single-phase induction generator based distributed power generation system. A prototype model is developed using MATLAB SIMULINK and tested in the laboratory under linear and non-linear loads. Based on implementation, the performance of control algorithm is validated and obtained results are found satisfactory.