Design and implementation of cloud connected photovoltaic monitoring system

The rising research and development studies in the domain of photovoltaic (PV) systems have made the PV systems one of the most well-known renewable energy sources. In particular, with invest supports, the use of small and medium-sized systems are increasing day by day. Although the costs of these systems are gradually decreasing, the initial investment costs are still high. Therefore, operating PV systems with high efficiency is very important in terms of shortening the depreciation periods. In this thesis, a cloud-connected monitoring system that can detect faults by using electrical and climatic data obtained from a PV system has been designed and implemented. For the thesis study, an experimental PV system was installed on a residential building in Muğla. Electrical and climatic data obtained from the experimental system by a data acquisition system which is developed, have been recorded by sending instantly to the cloud-computing layer and then used for fault detection. The proposed method for fault detection is based on the use of a machine learning method called ensemble learning, to classify faults in PV systems. The ensemble learning method aims to exceed the generalizability and prediction accuracy over a single machine-learning algorithm by combining the predictions of conceptually different learning algorithms. This method is also aforenamed as the committee of experts or the wisdom of crowds. The developed fault detection method estimates the operation state of the PV system by running on the cloud computing layer and then the prediction is presented as a response through a web service developed.