Small grid-connected photovoltaic systems up to 5 kW(p) are often not monitored because advanced surveillance systems are not economical. Hence, some system failures which lead to partial energy losses stay unnoticed for a long time. Even a failure that results in a larger energy deficit can be difficult to detect by PV laymen due to the fluctuating energy yields. Within the EU project PVSAT-2, a fully automated performance check has been developed to assure maximum energy yields and to optimize system maintenance for small grid-connected PV systems. The aim is the early detection of system malfunctions and changing operating conditions to prevent energy and subsequent financial losses for the operator. The developed procedure is based on satellite-derived solar irradiance information that replaces on-site measurements. In conjunction with a simulation model the expected energy yield of a PV system is calculated. In case of the occurrence of a defined difference between the simulated and actual energy yield, an automated failure detection routine searches for the most probable failure sources and notifies the operator. This paper describes the individual components of the developed procedure-the satellite-derived irradiance, the used PV simulation model, and the principles of the automated failure detection routine. Moreover, it presents results of an 8-months test phase with 100 PV systems in three European countries. (c) 2006 Elsevier Ltd. All rights reserved.