Hardware-in-the-Loop Platform with an Edge-Computing Node for Comparative Evaluation of Maximum Power Point Tracking Algorithms in Photovoltaic Systems

Maximum Power Point Tracking (MPPT) algorithms can perform properly in simulations but can degrade on real hardware because of delays, discretization, and computational limitations. This study presents a low-cost hardware-in-the-loop platform, with an MPPT controller runing on an edge-computing node, for the evaluation of MPPT algorithms of photovoltaic systems. A Raspberry Pi Pico 2W emulates both a photovoltaic module of three substrings with bypass diodes, which reproduces partial shading, and a boost converter whose model is integrated every 10 $\mathsf{\mu }\mathrm{s}$; a Raspberry Pi 5, acting as the edge node, runs the MPPT and a current loop every 5 ms. An improved Perturb and Observe (P&O) algorithm and a hybrid particle swarm optimization (PSO) algorithm, neither of which receives the irradiance, were compared on uniform and partially shaded step profiles with 10 seeded runs per algorithm and profile. Under partial shading, PSO achieved an energy efficiency of $91.9\pm 0.6\%$ vs. $82.4\pm 1.0\%$ for P&O and ended $90.9\%$ of the steps at the global maximum power point vs. $54.5\%$; under uniform irradiance, PSO obtained $94.7\pm 0.3\%$ and P&O $96.9\pm 0.1\%$. Efficiency is measured against the emulator model. The edge node held the 5 ms period with a standard deviation of 0.078 ms. Data and scripts are openly available.