SOC Profile Reconstruction from Residential EV Charging Data: An Optimisation-based Approach

The penetration of electric vehicles (EVs) in modern power systems is steadily increasing worldwide, highlighting the need to understand how EVs impact the power system. However, oftentimes the lack of detailed EV charging data, such as the EV state-of-charge (SOC), slows down the development of grid services using EVs. In this paper, we propose a methodology to reconstruct the SOC profile of EVs starting from their metered charging profiles. The methodology is applied to 11700 electric vehicle charging stations (EVSEs) from the Danish residential sector. The methodology consists of three stages: first, K-means clustering is adopted to remove EVSEs where the reconstruction is not possible, then, mixed integer linear programming is used to guess the EV driving consumption. Lastly, the entire SOC profile is reconstructure. Results show that the total energy stored in the analysed fleet (10000 EVs estimated) fluctuates daily in the 25-200 MWh range, with higher values recorded at 04:00 a.m., while the available daily storage capacity is up to 60 MWh. Results highlight that most EV owners already delay charging to periods where the electricity is cheaper. If the current EV market trend persists, by 2035, Denmark could have up to 20-30 GWh of available stored energy during the night, and 1.5-2.2 GWh of available storage capacity during the day, capable of storing 45-67% of the total current wind and solar production.