A hybrid supervised learning model for a medium-term MV/LV transformer loading forecast with an increasing capacity of PV panels

The share of photovoltaic (PV) generation has increased quickly in the last decade. Many PV panels are connected behind-the-meter (BTM), so that they can not be identified with measurement equipment at MV/LV transformers. This poses a challenge for a medium-term MV/LV transformer loading forecast if the capacity of PV panels is increasing over time. Therefore, this paper proposes a hybrid approach for a medium-term load forecast (MTLF) of a MV/LV transformer with an increasing capacity of PV panels that are not separately measured. This approach combines a supervised learning model (data-driven approach) with a model to estimate the generation profile of the PV panels (model-based approach). The results indicate that the accuracy of the forecast improves significantly, while an accurate generation profile of the PV panels connected BTM or a disaggregation of the net load is unnecessary.