Performance Evaluation of Regression-Based MPPT Algorithms Using Inverse SEPIC Converter under Partial Shading

NNumerous green energy resources, including solar, wind, bio, and hydropower, have garnered significant attention as effective alternative energy sources. Particularly beneficial to society and the economy, solar photovoltaic systems (SPVS) are the most preferred resource. Unfortunately, because of shadowing situations and fluctuating loads, these systems are unable to maximize power extraction under changeable irradiance. Many Lower Peak Power Points (LPPPs) and Global Peak Power Points (GPPPs) on their power voltage characteristics (P-VC) arise as a result of PSC. Therefore, these systems employ Maximum Power Point Tracking (MPPT) approaches. This work implements and experimentally evaluates two supervised learning MPPT schemes, Support Vector Regression (SVRT) and Linear Regression Based Technique (LRBT), for stand-alone photovoltaic systems under partial shading, using an inverse SEPIC converter. The main novelty is a hardware-aware, real-time evaluation of a computationally light LRBT MPPT on an inverse SEPIC topology, and a comparative analysis against SVRT on metrics relevant to practical deployment, including computational complexity, tracking time, output power or current, and tracking efficiency, under realistic partial shading conditions. Unlike prior ML studies that rely on simulation or heavy models, LRBT demonstrates fast convergence and very low computational cost suitable for microcontroller implementation. In MATLAB/Simulink experiments on a ۲×۲ PV array and inverse SEPIC converter, LRBT achieves a mean tracking efficiency of ۹۸.۳% (±۰.۲۵%), reduces tracking time to approximately ۰.۱۰ s (variance ۰.۰۰۰۸ s), and improves delivered power by about ۲.۰–۳.۰% relative to SVRT under the tested shading patterns. LRBT’s model size and prediction speed make it significantly more suitable for low-cost real-time hardware compared to SVRT.

کلیدواژه ها:

Linear Regression Based Technique ، Maximum power point tracking ، Partial shading Condition ، Solar Photovoltaic System MATLAB/Simulink