Advanced Failure Detection Algorithms and Performance Decision Classification for Grid-Connected PV Systems
Andreas Livera, Alexander Phinikarides, George Makrides, Juergen Sutterlueti, George E. Georghiou · EU PVSEC · 2017
The scope of this paper is to present the development of failure detection routines (FDRs) that will operate on acquired data sets of grid-connected PV systems and determine and classify accurately the exhibited failures. The developed FDRs comprise of a failure detection and a classification stage. Specifically, the implemented failure detection stage was based on a comparative algorithm that detected discrepancies between the measured and simulated electrical measurements (dc current, voltage and power of the array) of a test PV system. Furthermore, statistical algorithms for identifying outliers, anomalies and normal system operation limits were also used for failure detection. Accordingly, for each identified failure there was a subsequent decision stage which performed a classification process based on developed logic and decision trees. The decision trees were constructed with a supervised learning process, trained with continuous samples split in a 70:30 % train and test set approach of acquired data sets which included the feature patterns exhibited during normal and faulty operation. The results obtained by emulating three faults (shading, inverter and bypass diode failure), showed that the developed FDRs were capable of detecting accurately the faults upon their occurrence. Finally, when applying binary classifier to the test set which included several emulated failure patterns, the obtained results demonstrated that the classification accuracy was 98.7 % for the inverter failure, 95.3 % for bypass diode fault and 96.6 % for partial shading.