Use of a Digital Twin to Detect Stalling Tracker Events in Photovoltaic Power Plant with Horizontal Single-Axis Tracking Systems

Horizontal single-axis trackers are widely employed in photovoltaic power plants to enhance energy production. Despite their benefits, these systems encounter mechanical issues, such as stall events, which can lead to significant power losses. This work introduces a novel and broadly applicable methodology for detecting stalled tracker events through pairwise comparisons. The proposed algorithm operates directly on measured output power and requires minimal site-specific inputs. Validation was performed using both synthetic datasets – generated with physical models and randomized anomalies – and real field measurements from a photovoltaic array. Results from synthetic data demonstrate an F1 score of 0.60, highlighting potential for further refinement. Applying a day clearness threshold of 0.60 improved performance substantially, raising the F1 score to 0.96. When tested on field data under the same conditions, the algorithm achieved an F1 score of 0.75, pointing to additional challenges caused by phenomena that mimic stalled behavior.