This report summarises the exploratory data analysis and evaluation of basic forecasting models performed on data recorded with grid sensors located at MV-LV transformer stations in Chapelle–sur–Moudon site. These measurement data are provided by the distribution system operator Romande Energie. They consist of several power quality characteristics recorded at the LV side of 3-phase distribution transformers and on several distribution cabinets. The focus is on measurements of active and reactive power aggregated at MV-LV Champ Monet transformer station. Results of forecasting this transformer loading are also presented, applying CNNs, seasonal moving averages, and a weighted combination of CNN and naive models. Exploratory data analysis is conducted to extract representative load profiles for different seasons, and multiple forecasting models are evaluated using error metrics and visual error analysis. Highlighting the performance and stability of each method across seasons. These results provide insights into the temporal variations and predictability of transformer loading, setting the groundwork for integration into situational awareness workflows and dynamic tariff design.
