PV Polaris - Automated PV System Orientation Prediction

• Published in: IEEE photonics journal Vol. 17; no. 3; pp. 1 - 7
• Main Authors: Wittmann, Ernst, Buerhop-Lutz, Claudia, Bennett, Savannah, Christlein, Vincent, Hauch, Jens, Brabec, Christoph J., Peters, Ian Marius
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.06.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Automated orientation prediction; Automation; Azimuth; Inverters; Monitoring; monitoring data; Monte Carlo methods; Optimization; Orientation; performance optimization; Photovoltaic cells; Power generation; Prediction algorithms; PV location prediction; PV orientation measurement; PV polaris; Satellite imagery; Satellites; Temperature measurement; Vectors
• ISSN: 1943-0655; 1943-0647; 1943-0647
• DOI: 10.1109/JPHOT.2025.3568887

The orientation of a photovoltaic system is an important parameter for power generation and yield predictions. Yet often, the real orientation is unknown. Measuring the orientation manually is time-consuming. This study introduces an automated Monte Carlo Search based algorithm called PV Polaris which is capable of predicting the systems orientation within 18 s, with uncertainties of less than 2° in tilt and 4° in azimuth. In terms of accuracy, PV Polaris outperforms other methods such as measurements with a tilt compensated compass or predictions from satellite images. Applicable at module, string and inverter levels, the algorithm only requires power monitoring data as well as an approximate coordinate as input. Additionally, the algorithm can operate inversely to estimate the system's coordinates based on a given orientation. By using this orientation prediction, it was possible to calculate the yearly yield loss due to non-ideal orientation. For photovoltaic systems we investigated, we found that yearly yield increases between 2.3% to 10.3% could be achieved if the PV systems orientation would be optimized.