Diseño y simulación de un sistema MPPT basado en algoritmos de machine learning para autogeneración a pequeña escala fotovoltaica mediante un sistema embebido Raspberry Pi.

Abstract

An MPPT emulation and control system is presented, the objective of which was to design and simulate, using Raspberry Pi and MATLAB/Simulink, an emulated photovoltaic generator and a conversion stage connectable to a microgrid, posing as a central question how to implement maximum power point tracking in an embedded and simulation environment. The problem addressed was the variability of irradiance and the computational and sampling limitations that hinder real-time control and MPP identification under dynamic conditions. The objectives were to emulate the behavior of the PV generator on Raspberry Pi, model the conversion stage in MATLAB, select and train a machine learning algorithm for MPPT, and validate the Pi–MATLAB integration. These were achieved by implementing the diode model in Python on Raspberry Pi, simulating a buck-type DC–DC converter, a three-phase inverter, and a battery system in Simulink, and serial communication between the Pi and MATLAB for validation. For control, different MPPT methods were compared; the selected algorithm was ANFIS+IC, which was trained on 1 million data points and showed the best combination of accuracy and convergence time, achieving a reported median accuracy of 99.995%. In conclusion, the work demonstrates the feasibility of using Raspberry Pi as a model emulator/executor and of integrating advanced MPPT algorithms with MATLAB for small-scale applications.