Probabilistic Analysis of a Generalized Perturb and Observe Algorithm Featuring Robust Operation in the Presence of Power Curve Traps

Local maxima in the measured power curve of a photovoltaic (PV) panel, called traps, are caused by switching transients and other deterministic phenomena in PV power converters. Traps exist even in the absence of shading, and degrade the performance of perturb and observe (PO) maximum power point trackers. A generalized PO (GPO) algorithm is proposed which eliminates this degradation. PO/GPO algorithms in the presence of random noise are shown to be modeled by multidimensional Markov chains. These models quantify the tendency for PO at low random noise levels to become stuck in traps following insolation changes, thereby decreasing tracking performance and reducing harvested power. By contrast, GPO tracks insolation changes unhindered by traps. All theoretical results are experimentally verified using a hardware PV converter.