login

Search documents

Browse topics

Document details

 
Title:
 
Automatic Fault Detection and Classification in PV Systems by the Application of Machine Learning Algorithms
 
Author(s):
 
G.D. Rupakula, D. Daßler, S. Malik, M. Ebert, R. Schmidt
 
Keywords:
 
Defects, Monitoring, Performance, PV System, Classification, Inverter, Machine Learning
 
Topic:
 
PV Systems – Modelling, Design, Operation and Performance
Subtopic: Operation, Performance and Maintenance of PV Systems
Event: 38th European Photovoltaic Solar Energy Conference and Exhibition
Session: 5DO.1.2
 
Pages:
 
1011 - 1017
ISBN: 3-936338-78-7
Paper DOI: 10.4229/EUPVSEC20212021-5DO.1.2
 
Price:
 
 
0,00 EUR
 
Document(s): paper, presentation
 

Abstract/Summary:


For an effective functioning of photovoltaic (PV) plants, it is important to identify the potential faults (technical defects or yield loss due to harsh environmental conditions) that can occur during the operation of the plant and thereby derive possible maintenance work. This ensures a high energy yield from the PV plant throughout the operating time. This research work explains the approach to establish a full detection and classification model of a large-scale grid-connected PV power plant at the inverter level by the application of different machine learning (ML) algorithms. This work initially explains a statistical method to identify different faults at the inverter level based on the necessary model parameters which include module temperature, voltage and current at the inverter; solar irradiation and ambient temperature from a sensor. A machine algorithm and another statistical method are applied to this data for identification of normal functioning, fault detection and fault classification. Three important faults are detected in the present research namely, snow cover, inverter outage and degradation. An Artificial Neural Network (ANN) and a novel statistical method are proposed to detect these defects. The accuracy of the algorithms and the optimized parameters have been described and the possibility of integrating the algorithms into the monitoring platform of a PV power plant has been discussed.