login

Search documents

Browse topics

Document details

 
Title:
 
A Facility Test to Generate Data from Real PV Systems Affected by Faults and Validate Fault Detection and Diagnosis Algorithms
 
Author(s):
 
G. Maugeri, A. Rossetti, V. Urciuoli, D. Bombelli, S. Guastella
 
Keywords:
 
PV System, PV Monitoring, Fault Detection, Fault Diagnosis
 
Topic:
 
PV Systems Engineering, Integrated/Applied PV
Subtopic: Operation, Performance and Maintenance of PV Systems
Event: 8th World Conference on Photovoltaic Energy Conversion
Session: 4BV.5.10
 
Pages:
 
1364 - 1369
ISBN: 3-936338-86-8
Paper DOI: 10.4229/WCPEC-82022-4BV.5.10
 
Price:
 
 
0,00 EUR
 
Document(s): paper, poster
 

Abstract/Summary:


The use of Machine Learning (ML) techniques in various application domains--in computer safety systems or within industrial processes--has proven benefits from an economic and efficiency standpoint. With these methods applied to datasets from monitoring systems, in real-time, it is possible to reduce the time to recognize faults, restore normal operation quickly, and avoid serious or permanent malfunctions, which would lead to longer downtime and higher repair costs. In the photovoltaic (PV) sector, the automatic fault detection through algorithms employing ML approaches can provide important clues to maintenance operators for subsequent troubleshooting phase, enabling them to optimize repair time and costs. However, developing models for the detection and diagnosis of faults is far from trivial and numerous challenges must be faced. One of the most difficulties for the development, validation and optimization of Failure Detection and Diagnosis (FDD) algorithms, lies in accessing the historical production datasets of PV plants in which there are enough failure events for the training and tests of ML algorithms. Furthermore, the FDD algorithms need to be ready in handling data that may be affected by communication errors, uncalibrated sensors, or other factors that can affect the PV performances (shadings, grid interruptions, ecc). Within this work we are presenting an experimental Test Facility for the simulation of faults in PV systems, to generate large datasets from real PV systems that are affected by faults. A characterization of the main fault conditions that can occur in photovoltaic systems is presented, illustrating how the electrical parameter evolves in the specific fault. The real-field PV data generated by the RSE PV Facility Test allows having historical PV data with several separated failure events appropriately labeled for training and testing of ML algorithms. Within this work we are presenting an experimental Test Facility for the simulation of faults in PV systems, to generate large datasets from real PV systems that are affected by faults. A characterization of the main fault conditions that can occur in photovoltaic systems is presented, illustrating how the electrical parameter evolves in the specific fault. The real-field PV data generated by the RSE PV Facility Test allows having historical PV data with several separated failure events appropriately labeled for training and testing of ML algorithms.