Photovoltaic Inverter Input Arc-Fault and Normal Operation Waveforms Dataset
- Citation Author(s):
-
Michel Oliveira
(Mackenzie Presbyterian University (UPM))
Filipe Ramos
(Mackenzie Presbyterian University (UPM))
Gabryel Souza
(Mackenzie Presbyterian University (UPM))
José Neto
(Mackenzie Presbyterian University (UPM))
Bruno Lima
(Mackenzie Presbyterian University (UPM))
Lucia Saito
(Mackenzie Presbyterian University (UPM))
- Submitted by:
- Michel de Oliveira
- Last updated:
- DOI:
- 10.21227/wzfe-7973
- Data Format:
38 views
Abstract
The dataset contains voltage and current waveforms measured at the input terminals of a photovoltaic (PV) inverter during controlled arc-fault experiments.
The arcs were generated using a PV arc-fault generator powered by a solar battery through the inverter, and the corresponding signals were recorded using Tektronix DPO4034B oscilloscopes.
The measurements were performed under laboratory conditions to simulate different arc behaviors and to support the development of machine-learning and signal-processing techniques for arc-fault detection in photovoltaic systems.
Instructions:
The corpus comprises 16 experiments (Experiment_01 … Experiment_16). To provide realistic operating context, seven experiments are baseline “no-arc” runs that only contain small, natural variations in DC current and voltage (load/irradiance-like drifts) without intentional arcing: 01, 02, 05, 09, 10, 13, 14. The remaining nine experiments contain pronounced arc-fault events. This balance enables both supervised and unsupervised benchmarking.
Files and formats
- CSV (16 files): Each file has 1,000,000 samples acquired at 4 µs sample interval (~4 s duration) with columns TIME, CH1, CH2, CH3 plus Tektronix metadata headers (model, firmware, record length, probe attenuation, vertical units/scale/offset, etc.).
- CH1: DC current (A) at inverter input.
- CH2: DC voltage (V) at inverter input.
- CH3: Unused.
- TIME is the time axis (s), spanning pre-trigger and post-trigger.
- PNG (16 files): Oscilloscope screenshots corresponding 1-to-1 to the CSVs, visualizing CH1 (current) and CH2 (voltage) during each run.
All files are provided in open formats (.csv, .png) for straightforward use in Python, MATLAB, or R. The dataset is intended for studying transients, building feature-extraction pipelines, and training/evaluating detectors of PV arc-faults under both event and no-event conditions.
ETHICS AND LICENSING
- No human subjects; only instrumental electrical signal data.
- Please cite the associated IEEE Access manuscript and this dataset when using any part of the data or derived artifacts.
