Digital twin-driven surrogate optimization for maximum power point tracking of photovoltaic systems under partial shading conditions

Photovoltaic systems under partial shading develop multimodal power-voltage characteristics that render conventional gradient-based MPPT algorithms ineffective. Bio-inspired metaheuristic approaches overcome local-maxima entrapment but introduce a structural scan-phase energy loss: each search iteration physically evaluates a candidate duty cycle on the PV hardware, dissipating real energy at every shading event. This paper presents the Digital Twin-Driven Surrogate-Optimized MPPT (DTSO-MPPT) model, which substantially reduces hardware scan losses by decoupling the global search from the physical PV system and limiting hardware interaction to VTC-based verification. The proposed four layer architecture has a physics calibrated Digital Twin with Recursive Least Squares, a surrogate PV surface builder to evaluate the calibrated model, an Orangutan Optimization algorithm to search the surrogate for the global optimum and finally a validation and transfer controller to verify the accuracy conditions. The proposed model was implemented and numerically evaluated in MATLAB R2024b/Simulink 24.1 under five partial-shading conditions. The simulation results exhibit the proposed model tracking efficiency as 99.1%, GMPP convergence in 0.14s, steady-state power oscillation averaging approximately 0.53 W and remaining within 0.42–0.58 W across the five evaluated cases. The scan-loss reduction energy efficiency metric of 99.4% obtained by the proposed quantifies the reduction in scan-loss, which represents the principal advantage over conventional hardware-coupled MPPT methods.