Independent component analysis for harmonic source identification in electric power systems

In the last couple of decades harmonics have become a major power quality problem in electric power systems. It is important to identify the harmonic sources in the system to solve and prevent harmonic related problems. It is the original contribution of this thesis to model the harmonic load identification problem as a blind source separation task and to solve it using a statistical technique called Independent Component Analysis (ICA). Under non-sinusoidal conditions, harmonic voltage measurements are modeled as a linear combination of statistically independent harmonic current sources. This thesis demonstrates that ICA is well suited to estimate the harmonic current sources using a relatively small number of measurements and without knowledge of network topology and parameters. In addition to the harmonic source estimates, ICA also provides a rough estimate of the system admittance matrix. This matrix provides information for the location estimation of harmonic sources. A discussion of the sensitivity of the estimation algorithm with respect to measurement noise, number of data points and measurements, and the statistical independence is presented. The effectiveness of the proposed estimation algorithm is demonstrated through computer simulations using several ICA methods and power system test cases.