Photovoltaic Maximum Power Extraction and Delivery based on Adaptive PSO-Neural Network-Fuzzy Algorithms Implemented using an Experimental Test Bench

: This paper reports the hardware implementation of two novel Maximum Power Point Tracking strategies for photovoltaic systems. To maximize harvested energy, the photovoltaic source must be regulated for high efficiency while keeping the controller low-cost, computationally light, and easy to implement. It is proposed that two Maximum Power Point Tracking algorithms that use Adaptive Particle Swarm Optimization to train a neural network. The first, based on Adaptive Particle Swarm Optimization, enhances the classic fixed-step Perturb-and-Observe method; the second integrates a fuzzy-logic regulator. Both approaches are designed to overcome the limitations of conventional Maximum Power Point Tracking techniques. The controllers drive a non-isolated, high-gain DC/DC boost converter used as an impedance adapter to optimize power transfer from photovoltaic modules to the load. Simulink models were developed and benchmarked against standard Perturb-and-Observe and fuzzy-logic methods, with simulations indicating clear benefits for the Adaptive Particle Swarm Optimization-based neural network with fuzzy-logic regulation. A hardware test bench—comprising a solar emulator and a dSPACE 1104 board—was then built to validate performance. Experiments confirm that the Adaptive Particle Swarm Optimization-based neural network with fuzzy-logic regulation achieves fast execution (0.03 s), high robustness, no steady-state oscillations (steady-state error < 10 -5 ), and no measurable energy loss, with a measured overshoot of 19.5% across diverse environmental and load conditions.