We present a remote diagnostic method that uses the IV measurement function of smart inverters and a so called self-referencing procedure to represent the performance of the PV generator vs. operating conditions irradiance and temperature (G and T). We have recorded 200 IV-curves of a PV string within a period of six months using a smart inverter. A deep autoencoder detected disturbed or IV measurements so that these were not included in the evaluation. The effective irradiance Geff at the PV string was determined, which was included in the evaluation instead of the measured irradiance. In addition, the cell temperature of the PV generator Teff was determined using physical models for the evaluation. As a result, we create smooth power-surfaces over Geff and Teff conditions in the range of 100–1100 W/m² and 15–90 °C. For validation, the performance data of the PV string were compared by indoor measurements with a calibrated flasher at standard test conditions. The approach offers a remote and real-time diagnostic by smart inverters. It is well suited for accurate power monitoring of PV generators or degradation or soiling tracking without the need for additional sensor capabilities. 

Keywords: Smart inverter IV tracing, IEC 61853-1, G–T performance matrix, Degradation monitoring, Soiling 

In this work, the factors leading to string outages are examined, and an enhanced method for detecting faults at the substring

level is presented. Utilizing GPT4-o to analyze O&M reports of 5089 photovoltaic plants, we classified outages

according to the affected components and the underlying origin, identifying the most frequent string fault causes. An

approach employing CUSUM Charts is introduced to identify substring outages within PV systems effectively. The methodology

utilizes fundamental field data that is commonly available in practice. A filtering approach, combined with the use

of CUSUM control charts, minimizes false positives, ensuring that only consistent underperformance is flagged as an outage.

The methodology returns far fewer false positives and more stable error intervals for substring outages than a former

monitoring approach. Overall, the study demonstrates a significant improvement in detecting substring outages. The

advanced methodology enables more effective O&M for PV plants, where substring outages are reliably identified after a

short detection time.

Die präzise Erkennung von Clear-Sky-Momenten ist für die Überwachung und Effizienzana-lyse von Photovoltaikanlagen von zentraler Bedeutung, da zu diesen Zeitpunkten definierte und model-lierbare Einstrahlungsverhältnisse herrschen. Es wird ein hybrides Modell zur verbesserten Erkennung von Clear-Sky-Momenten auf Basis von Einstrahlungsdaten vorgestellt. Hierfür wurden zunächst ma-nuell, dann mithilfe eines CNNs Merkmale aus den Einstrahlungsdaten gebildet. Eine Falls tudie mit Referenzdaten belegt, dass durch die Kombination dieser wissens-und datengetriebenen Methoden Clear-Sky-Momente zuverlässiger identifiziert werden können. Dadurch können Analysemethoden schneller und zuverlässiger Aussagen über die untersuchten PV-Anlagen treffen.

On-site current-voltage (IV) measurements will play an essential role in the online monitoring of PV systems. However, challenging measurement conditions like inconsistent irradiance levels on PV arrays (e.g., due to local shading) can distort IV curves, leading to inaccurate characterizations. By accurately detecting deformed IV curves, the reliability of both on-site and remote IV measurements is significantly enhanced. For this purpose, several classifiers were evaluated using 4104 manually labeled IV measurements on a mc-Si-PV array. Machine learning tech-niques perform much better than a traditional rule-based filter, with accuracy above 99 %. A deep Autoencoder was employed to reduce IV measurements into a set of 7 features, which encoded the shape of the curves into a low dimen-sionality. The IV-Autoencoder improved the classification of IV curves, yielding better results than a feature reduction with Principal Component Analysis. The proposed classifiers are able to sort out on-site IV measurements under un-satisfactory environmental conditions, benefiting the online monitoring of PV systems. It may also be used as an indi-cator for faulty PV strings.