Abstract
This paper proposes a reduced sensor-based nonlinear maximum power point tracking (MPPT) controller for grid-integrated photovoltaic (PV) systems operating under rapidly changing climatic conditions. Unlike conventional approaches that require costly irradiance sensors, the proposed method employs a mathematical irradiance estimation model and a radial basis function neural network to generate optimal reference voltages, which are then enforced by a backstepping nonlinear controller. This two-stage design enables fast and robust MPPT while maintaining DC-link stability and grid power quality. The controller was validated on a 100 kW MATLAB/Simulink-based grid-tied PV system with a DC-DC boost converter and inverter. Under step changes in irradiance, the system tracked the new MPP in as little as 7 ms, while restoring DC-link stability (500 V) within 42 ms. Under continuously varying conditions, it maintained synchronization with the grid and achieved a total harmonic distortion (THD) below 0.1%. Comparative results against Perturb & Observe (P&O), Improved Differential Evolution (IDE), and Particle Swarm Optimization (PSO) demonstrated that the proposed method achieved the highest PV-side power yield (80.41 kW vs. 79.71 kW for P&O, 73.44 kW for IDE, and 59.34 kW for PSO), the highest grid-side active power delivery (78.69 kW vs. 77.98 kW, 72.14 kW, and 58.29 kW respectively), and the lowest integral absolute error of DC-link voltage (IAE = 10.6156). These results confirm that the controller provides faster convergence, improved voltage regulation, and superior grid stability compared to state-of-the-art MPPT methods, making it a promising solution for real-world deployment in large-scale PV systems.