Blind signal separation of harmonic voltages in non-linear loads

Abstract EN

Extensive use of nonlinear loads like arc furnace and PWM Inverters generate considerable voltage harmonics.

These harmonics need to be estimated and quantified using minimum knowledge about the topology of the system.

Blind Signal Processing Techniques like Fast ICA (FICA), Joint Approximate Diagonalisation of Eigen Matrices (JADE) and Entropy Bound Minimization (EBM) are applied for harmonic voltage estimation in a simple four bus system.

Conventional ICA algorithms like FICA and EBM algorithm break down when only the arc furnace load contributes to the harmonics in the power system.

An algorithm termed as FICOMB which is a combination of Fast ICA and COMBI is employed for harmonic voltage estimation.

The graphical results and error indices of a simple four bus system indicate that FICOMB and JADE are suitable for harmonic frequencies like h = 3 and EBM is suitable for h = 5 The primary objective of the present work is to use the estimation further for either the design of mitigation equipment or for identifying the perpetrators of harmonics with accuracy being more significant for both the applications.

This calls for an optimal estimator to choose a definite algorithm for a particular frequency.