Adaptive Gradient Descent MPPT Algorithm With Complexity-Aware Benchmarking for Low-Power PV Systems

Abstract

This paper proposes a computationally efficient, real-time maximum power point tracking (MPPT) algorithm tailored for low-power photovoltaic (PV) systems operating under fast-changing irradiance and partial shading conditions (PSC). The proposed method augments the classical perturb and observe (P&O) algorithm with an adaptive gradient descent mechanism that dynamically scales the perturbation step size based on the instantaneous power-voltage slope, thereby minimizing tracking time and steady-state oscillations. An optional initialization routine enhances global MPP (GMPP) tracking under PSC. Extensive simulations, including irradiance recordings from freely moving rodent subjects relevant to the targeted application, and tests across varying converter topologies and temperatures, demonstrate its robust, topology-independent performance. The proposed algorithm achieves 99.94 percent MPPT efficiency under standard test conditions (STC), 99.21 percent when applied to experimental data, and more than 99.6 percent for the tested temperature profiles. Under PSC, the initialization routine improves tracking efficiency by up to 7.8 percent. A normalized gate-level complexity analysis and a unified figure-of-merit (FoM) incorporating efficiency, tracking time, and computational cost demonstrate that the proposed algorithm outperforms 35 state-of-the-art P&O-based MPPT algorithms. These results underscore its suitability for integration in low-power power management integrated circuits (PMICs) operating under dynamic and resource-constrained conditions.