Synne Krekling Lien, Lasya Priya Kotu, Åse Lekang Sørensen, R Jayaprakash
Electric vehicle (EV) charging at home is transforming the electricity load profile of residential buildings. Using data-driven methods to extract information about EV charging can provide valuable insights for grid management, energy planning and demand side flexibility to avoid increased peak loads in the grid with the electrification of the transport sector. Research has shown that detecting EV charging in smart meter data can be quite successful in households with few other electric loads, as EV charging often dominates peak loads. However, little research has addressed the disaggregation of EV charging in households with electric heating, where load profiles show large variations. To address this gap, this paper combines real household smart meter data from 296 single-family houses in Norway with and without electric heating, together with EV charging data from 82 residential chargers to create artificial load profiles from households with and without EV charging. It presents a three-step method for (1) estimating EV charger capacity from household smart meter data, (2) classifying charging events (on/off) from smart meter data, and (3) estimating electricity consumption for EV charging from smart meter data. The results show that Step 1 estimated EV charger capacity with high accuracy using smart meter data, achieving 93% accuracy with classification and R² of 0.91 with regression. Step 2 detected charging events with 87 % precision with Categorical Boosting classification with (CB). Step 3 estimated charging load using regression, achieving R² of 0.80–0.86 and NMAE of 0.005–0.006, while simple multiplication of results from Steps 1 and 2 performed poorly. Each step can be used independently for applications such as grid planning, user nudging, and EV load disaggregation.