Arc detection in DC network-oriented identification by means of compressed sensing and machine learning.
Since they are direct current (DC) systems, photovoltaic installations carry the risk of electrical arcing due to poor connections. Such poor connections arise due to a variety of reasons such as faulty or aging connectors, human error, and environmental influences. Electrical arcing in a photovoltaic system can result in a reduction of output energy. More seriously, arcing can cause fire, which may in turn result in property damage and/or serious injuries.
Existing safety devices to combat electrical arcing are often optimized for specific applications, and suffer from a high rate of false detection that result in needless and costly system shutdowns. The aim of FlashCheck is to provide a general solution to detection of electrical arcing in photovoltaic systems while avoiding false detection. In order to accomplish this, a database of electrical arc signatures will be collected using compressed sensing, with the aim of deploying machine learning techniques to characterize arcs to enable future detection.
The RISC Software GmbH is involved in the development and deployment of compressed sensing techniques to reconstruct high-frequency samples with a low sampling rate.
FlashCheck is a cooperative research and development project supported by the Austrian Research Promotion Agency (FFG). The project is headed by Fronius International GmbH, with project partners FH Oberösterreich Forschungs und Entwicklungs GmbH, RISC Software GmbH, and Eaton Industries