Adaptive Estimation of Higher Order Harmonics of Power Systems Using Gradient Algorithms and Robust Model Based Learning
Harmonic distortion affects the quality of control of power grid, and estimation of harmonic distortion is an important task for controlling of power systems. Harmonics are caused by nonlinear loads such as power inverters, variable speed drives, and other inductive loads. Harmonics decrease the quality of power and can cause problems for power systems such as equipment damage, high loading, inaccurate metering, outages, and system losses.
I have studied various adaptive estimation algorithms in order to estimate the harmonic contents of a given signal. I am proposing a gradient, modified gradient, and robust model based learning algorithms to estimate the amplitudes and frequencies of sinusoidal harmonics. The algorithms are verified with sampling signals from a real voltage inverter. The algorithms have been proven to significantly simplify the modeling and simulation of converters.
Simulations are run using a simulated and a real-world signal from a voltage source inverter. The simulations are run with varying known and unknown parameters in order to display our findings.
Through this research I am able to show that it is not necessary to know every component of the system in order to perform analysis on the harmonic distortions. With a sample output signal, one can correctly model the harmonic contents of a system and inject them into an average model in order to model a system without all of the components.