PV-Fault-DS226: A High-Resolution Multivariate Dataset for Photovoltaic Fault Diagnosis in Tropical Climates

PV-Fault-DS226 is a high-resolution multivariate dataset for photovoltaic (PV) fault diagnosis and prognosis under tropical coastal conditions. It was collected from a dual-panel experimental testbed (NJR-120P-36 and Win Bright YB-156P36-120 modules) installed in Douala, Cameroon (4°02'33.8" N, 9°41'32.3" E), over an 11-day outdoor measurement campaign.

The dataset comprises 226,175 synchronized observations of terminal voltage (V1, V2), output current (I1, I2), module temperature (T1, T2), and plane-of-array irradiance (Irr), acquired via a Raspberry Pi 3 hub with a 16-bit ADS1115 analog-to-digital converter, ACS712 Hall-effect current sensors, Pt100/MAX31865 temperature probes, and an RS-RA-N01-AL-EX irradiance sensor.

Five operating conditions are physically represented, using Hardware-in-the-Loop fault induction (Schneider Easy9 circuit breakers and Bourns precision potentiometers) rather than software simulation:

• Label 0: Healthy operation
• Label 1: Short-circuit
• Label 2: Open-circuit
• Label 3: Progressive resistive degradation (aging/insulation loss)
• Label 4: Partial shading

Files included:

• df_ia_raw_acquisition.csv file synchronized multivariate acquisition data (voltage, current, temperature, irradiance)
• PV_Diagnostic_Final_Labels.csv (ground-truth fault classification per observation)

This dataset is released under a CC-BY 4.0 license: reuse is permitted provided appropriate credit is given to the authors as shown in the citation field of this record.