Accelerating Photovoltaic System Simulations via Statistical Data Aggregation

Accurate energy yield assessments (EYAs) for photovoltaic (PV) systems are critical for their financial viability, yet simulating the non-linear effects of partial shading over a full year remains computationally prohibitive. High-fidelity models, essential for capturing these mismatch losses, create a significant bottleneck in the rapid design and optimization of PV projects. This ieco.io work introduces a novel data-centric framework to accelerate annual PV simulations by applying statistical aggregation, reducing the input dataset of operating conditions while preserving the high-resolution information necessary for accurate shading analysis. We propose and evaluate two methods based on k-means clustering, with results clearly demonstrating the superiority of one approach. A key advantage of this method is its tunability, enabling substantial reductions in simulation time while introducing only minimal, controllable error, thereby outperforming standard module-level simulations in both speed and accuracy. Importantly, when quantifying annual shading-induced power losses, the method drastically reduces computational overhead with negligible impact on accuracy, contrasting sharply with the significant underestimation of losses inherent to the module-level approach. This framework offers engineers a powerful and flexible tool for fast, reliable energy yield assessments without compromising simulation fidelity.