Machine Learning Methods for Partial Shading Detection in Monitoring Data on PV-Systems

The focus of this work is to demonstrate the advantage of collecting PV-power system monitoring data with modern communication technology and the intelligent data analysis algorithms of computer science. Prediction and recognition of faults in large PV-arrays are very important for the effective operation of PV-plants. Different PVSystems are monitored with a large amount of data collected. The challenge in this work is using intelligent data analysis methods of machine learning algorithms for a fault detection in PV-plants. The focus is on partial shading recognition for a laboratory test site and industrial PV-plants. Recognition rates are calculated for laboratory PV-test site, where information on electrical and environmental data is given in high resolution and correctness and for field installations, where information is given on module voltage and string power every 10 minutes. Temperature values per module where also taken into account.