Maximizing solar energy harvesting efficiency: Optimal hybrid deep neural learning - based MPPT for Photovoltaic systems under complex partial shading conditions

• Published in: Sustainable computing informatics and systems Vol. 47; p. 101159
• Main Authors: SeyedShenava, SeyedJalal, Zare, Peyman, Davoudkhani, Iraj Faraji
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.09.2025
• Subjects: Artificial neural network; Complex partial shading; DC/DC converter; Deep learning; Incremental conductance; Maximum power point tracking; Perturb and observe; Photovoltaic systems; Photovoltaic systems; Deep learning; Maximum power point tracking; Incremental conductance; Artificial neural network; Complex partial shading; Perturb and observe; DC/DC converter
• ISSN: 2210-5379
• DOI: 10.1016/j.suscom.2025.101159

The declining viability of fossil fuels and their adverse environmental impacts are accelerating the global transition to Renewable Energy Sources (RESs), with solar energy emerging as a key pillar due to its versatility and scalability. Photovoltaic (PV) systems enable direct solar-to-electric conversion but face challenges such as nonlinear behavior and multiple Local Maximum Power Points (LMPPs) under Complex Partial Shading Conditions (CPSCs). This study introduces an enhanced Maximum Power Point Tracking (MPPT) method based on a hybrid Artificial Neural Network–Improved Incremental Conductance (ANN-IINC) model. The ANN is trained using representative datasets capturing diverse shading patterns to estimate optimal reference voltages dynamically, while the IINC module accelerates convergence with reduced oscillations. To validate the proposed method, three CPSC scenarios are simulated and compared with traditional perturb and observe and INC techniques, as well as recent metaheuristic optimization algorithms. Sensitivity and descriptive statistical analyses confirm that the ANN-IINC approach not only achieves faster convergence (81.9 ms) and higher tracking accuracy (up to 99.9096 %) but also reduces standard deviation in power output by 11.3 %–14.8 % compared to classical methods. Furthermore, confidence intervals for efficiency are narrowed by over 20 %, demonstrating improved robustness and statistical significance. The method's computational complexity is optimized, maintaining real-time applicability without sacrificing precision. A comprehensive adaptive analysis and hyperparameter sensitivity study further reinforce the superiority and practical relevance of the hybrid architecture. The study offers a scalable, stable, and efficient solution to the MPPT problem under dynamic environmental conditions. These results highlight the ANN-IINC technique’s capacity to outperform both classical and metaheuristic MPPT strategies, contributing meaningfully to the advancement of intelligent PV control under CPSCs. •Introducing the groundbreaking IINC-ANN hybrid method to optimize MPPT efficiency under CPSCs.•Thorough evaluation of PV system performance under CPSCs via meticulous simulation analysis across diverse solar radiation patterns.•Comparative assessment against P&O and conventional INC techniques, revealing the IINC-ANN hybrid's superiority in handling CPSCs challenges.•Remarkable enhancements in system efficiency up to 99.9378 %, 99.9096 %, and 99.4883 % with the adoption of the hybrid approach.