The photovoltaic (PV) systems play an increasingly important role in the power systems. PV production of some countries have already achieved a significant penetration rate. In 2015 the PV production of Italy meant 8% of the whole consumption of electricity, in Greece, Germany and Belgium this rate was 7,4%, 7,1%, 4%. This data represents the significant of the PV production refer to the plan and operation of power system.
The forecast of PV production is relevant regard to preventive actions and reserve allocation of TSO (Transmission System Operator). For the predictions different input data is needed such as satellite pictures, numerical weather prediction data, which usually use machine learning algorithms, statistical or physical methods.
Furthermore, prediction of PV production is also important for intraday market of Hungarian mandatory off-take (KÁT). Day ahead market timetable can be made accurate with modification of intraday timetable during the day. Less balancing activity for TSO effect higher level of system security and decreasing amount of balancing energy, moreover, in a longer term, the need for system reserves may decrease.
In addition, forecasting of photovoltaic production is also relevant for monitoring and regulation of low voltage networks.
Previous aspects can reveal to the important of qualitative examination of PV production. In this thesis I would like to present the methods regarding to production of PV plants and concrete examinations such as k-means clustering and prediction using neural networks.