Multi-stage day-ahead scheduling for building-integrated large-scale electrical vehicle charging station

Abstract

Electric vehicle (EV) fleets have grown significantly over the past decade. To meet the increase in EV charging demand, one scenario considered is to convert parking spaces in functional buildings into EV charging stations, so that the large-scale EV charging demand is added to the building microgrid in a cooperatively controlled manner to avoid the negative effect on the building itself as well as on the main grid. In this work, we consider a building charging microgrid (BCM) system equipped with a solar photovoltaic (PV) system and a stationary battery-based energy storage system (BESS), which aims to achieve three objectives: 1) minimize their operating cost, 2) commit to acting as a well-behaved load in the day-ahead electricity market by flattening their power profile and following their day-ahead schedule; and 3) balance PV generation, given the limitation of the main grid, the capacity/power limit of the BESS, and the stochastic behavior of the EV parking pattern, building load, PV generation, and electricity price. The biggest challenges faced by BCM operators are the highly stochastic behavior of the EV parking models, the high computational complexity, and the dynamics of EV load. To address these challenges, we propose a multi-stage charge/discharge scheduling framework, which consists of three main steps: 1) a day-ahead scheduling optimisation is solved one hour in advance of the operating day; 2) an intra-hour rolling horizon optimization is solved every 15 minutes adapting to the forecast errors of EV demand; 3) a real-time heuristic control and adjustment algorithm is based on the laxity and stage of charge (SOC) of the EVs and their real-time flexibility assessment. A stochastic modeling approach for EV fleets adopting the Monte Carlo simulation method assuming known probability distribution of the arrival time, parking duration, initial SOC, battery capacity, and diversity of the charging rate is proposed to capture the stochastic nature of EV behaviors. To reduce the computational complexity related to large-scale EV fleets, a novel cluster-based aggregation technique which divides EV fleets into three clusters based on their laxity and discharge capability is applied to guarantee high satisfaction of EV users. Finally, a real-time adjustment algorithm is applied to track the day-ahead power schedule. The performance of the proposed algorithm is measured through extensive simulations in office, residential and commercial BCMs using both real and simulated data. Our simulation results show a 4.42% reduction of operating cost in average while maintaining customer satisfaction at over 97% thanks to the real-time electric vehicle flexibility assessment and the new cluster-based aggregation model.

Date	4 Jan 2022
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Kamal Al-Haddad (Supervisor) & Kim Khoa Nguyen (Co-supervisor)