Worldwide solar energy production increased in the recent years with the rapid population of the rooftop PV systems. This growth caused operational and planning problems due to the uncertainty of the solar power-based generation. Therefore, proper monitoring of solar generations is critical for distribution system operators, as the operators use solar generation data collected from the field, to make accurate forecasts. In case of a malfunction, the system operator should be aware, and make the required modifications in the forecast process. On the other hand, PV system owners also require monitoring of their systems to take action in case of a malfunction. Considering that number of operational PV systems is very high, and occurrence of a malfunction is rare, use of automatic malfunction detection methods necessitates. This paper develops a method that detects any malfunction of a PV system. The malfunction may refer to failure of communication network, sensor malfunctions or data logging unit errors as well as problems of inverters and PV panels. The proposed method employs Deep Feed Forward Neural Network to predict solar generation data based on current weather conditions and solar radiation measurements, and compares the prediction with the measurements using the normalized residuals test. The proposed method is tested using real field data recorded at the Ayasli Research Center of the Middle East Technical University.