An interpretable multi-level classification decision-based maximum power point tracking method for photovoltaic systems under partial shading conditions

In photovoltaic power generation systems, artificial intelligence technology is widely utilized in areas such as power forecasting, photovoltaic fault diagnosis, and global maximum power point tracking. The nonlinear variation properties of photovoltaic modules in different environmental conditions cause photovoltaic power loss. Meanwhile, photovoltaic systems face the global maximum power point tracking issue under partial shading conditions. Therefore, this study proposes a global maximum power point tracking method based on interpretable multi-level classification decisions (IMCC). The proposed IMCC method establishes an interpretable classification network that divides duty cycle data into various categories. Meanwhile, the IMCC method utilizes a classification decision mechanism divided into intervals and stages to refine the duty cycle change amount. Hence, the proposed IMCC method can realize precise global maximum power point tracking through multi-level classification decisions. In addition, the interpretability of the IMCC method provides transparent and reliable assurance for the control of photovoltaic output power. Finally, this study conducts simulation tests and experimental verification of the IMCC method compared with traditional methods and swarm intelligence algorithms. The experimental results reveal that the IMCC approach increases tracking efficiency by 0.474 % and reduces steady-state oscillations by at least 6.016 W compared to the contrast algorithm.