Adaptive gradient descent–based maximum power point tracking algorithm for low-power photovoltaic systems with complexity-aware performance analysis

Maximum power point tracking (MPPT) in low-power photovoltaic (PV) and optical wireless power transfer (OWPT) systems must simultaneously provide high tracking efficiency, fast response, and low implementation cost. However, conventional perturbative methods often suffer from slow convergence or steady-state oscillations under rapid irradiance changes, whereas intelligent and metaheuristic approaches can improve tracking but typically increase computational complexity, memory demand, and hardware cost, limiting their suitability for compact embedded energy-harvesting systems. Here, we propose an adaptive gradient-descent-based perturb-and-observe MPPT algorithm for resource-constrained implementation. The method uses the analytical power–voltage gradient to adapt the duty-cycle perturbation direction and step size, enabling fast transient tracking, while a threshold-based suppression mechanism reduces steady-state perturbations near the maximum power point. A lightweight optional initialization routine based on a coarse duty-cycle sweep improves global tracking under partial shading with limited memory and computational overhead. We also introduce a normalized ASIC-oriented computational-complexity analysis for operation-level comparison of conventional, intelligent, and hybrid MPPT algorithms. Evaluated in a fixed-point MATLAB/Simulink framework with two 10-bit ADCs, an 8-bit duty-cycle command, a DC–DC boost converter, and a 1 MHz PWM stage, the proposed algorithm achieves 99.94% MPPT efficiency under standard test conditions, 99.21% under experimentally derived motion-induced irradiance, and up to 7.8 percentage-point improvement under severe partial shading.