Dissertation · 2026
Modeling and Analysis of the Operational Characteristics of Solar Panels in Photovoltaic Power Plants
Solar power plants often monitor performance at the inverter or string level, making degradation in an individual panel difficult to identify. This research explores panel-level anomaly detection using operational telemetry collected from photovoltaic systems.
The proposed approach combines physics-informed digital twins, long-term statistical analysis, and autoencoder neural networks. These methods estimate expected panel behavior, identify persistent deviations, and detect unusual telemetry patterns that may indicate degradation or potential equipment failure.
Research at a glance
Operational telemetry
Voltage, current, temperature, and irradiance measurements
Three analytical approaches
Digital twins, statistical analysis, and autoencoder networks
Panel-level analysis
Identifying abnormal behavior hidden within aggregate measurements
Methods
Physics-informed digital twin
Models the expected electrical behavior of individual solar panels using operational and environmental measurements.
Long-term statistical analysis
Examines persistent and changing performance deviations across extended periods of telemetry data.
Autoencoder neural network
Learns normal telemetry patterns and identifies panels whose behavior differs from expected operating conditions.
Key findings
- The research demonstrated that individual panel anomalies can be identified even when aggregate plant measurements appear normal.
- Combining physics-based models, statistical analysis, and neural networks provides complementary perspectives on equipment behavior.
- Long-term telemetry analysis can reveal gradual degradation that may be missed by short-term monitoring.
- Autoencoder networks can identify unusual patterns that are not always captured by predefined physical or statistical models.
- Selected anomalies were further investigated through technical and laboratory analysis.
Research output
The research resulted in journal publications, international conference papers, and a registered software work related to photovoltaic monitoring and anomaly detection. 14 published scientific works: peer-reviewed journal articles and international conference papers on AI-based anomaly detection in solar energy systems.