The energy domain represents a part of critical infrastructure whose influence has been growing significantly in recent years. Successful modelling of photovoltaic energy production enables more efficient management and planning of energy production from various energy sources (wind, nuclear, solar, geothermal energy, ...). More efficient planning can thus positively affect energy savings as well as lower production costs. Successful modelling of photovoltaic production requires monitoring of selected primary meteorological attributes which significantly affect the production rate. The aim of this article is to define derived attributes, enabling successful modelling of photovoltaic energy production across locations with different meteorological characteristics. This significantly simplifies the modelling process and increases the accuracy of trained regression models. The subsequent validation confirmed the high accuracy of the implemented modelling using machine learning regression models.