Benchmarking Machine Learning Architectures for Direct Maximum Power Point Tracking Control in Photovoltaic Systems

Maximum Power Point Tracking is vital for maximizing energy extraction in photovoltaic systems. Traditional methods, such as Perturb and Observe and Incremental Conductance, often suffer from low efficiency and steady-state oscillations. To address these limitations, this study evaluates numerous machine learning architectures using the MATLAB Regression Learner toolbox. Distinctively, load resistance is integrated as an auxiliary input alongside irradiance and temperature to ensure tracking independence from external controllers or algebraic dependencies. Results demonstrate that ML models achieve approximately ۹۹% efficiency with ripple-free power output. These advancements confirm ML as a robust framework for high-performance solar energy harvesting.

کلیدواژه ها:

MPPT ، Machine Learning ، Photovoltaic Systems ، MATLAB Regression Learner ، Solar Energy