Data Driven Stochastic Energy Cost Optimization with V2G Operation in Commercial Buildings

Abstract

Electric vehicles are being considered as a way of providing a sustainable transportation system. The bidirectional capability of any Plug-in Electric Vehicles (PEV) help in optimal overall energy use and demand reduction. This paper presents a novel data driven stochastic optimization method for energy cost reduction of a typical commercial building located in Southern California region. Uncertainty of PEV availability for charging and discharging implies stochasticity to the regular energy cost optimization problem for the consumers behind the meter. A probabilistic model is formulated by using a Mixed Integer Linear Programming (MILP) to minimize the overall energy cost. One-year actual data of light-duty PEV are collected which contain information regarding PEV availability, initial SOC level, change in SOC level, count of charging events, etc. along with total distance travelled. The optimization model developed incorporates all these detailed components representing real world scenarios. Multiple probabilistic distributions on PEV availability have been tested to show the impacts of PEV randomness on expected energy cost reduction. The impacts of PEV owner strategies are also shown to address the willingness of the PEV owner for participating in Vehicle to Grid (V2G) programs. Acceptable SOC strategy which reduces range anxiety at the end of PEV plug-in activity always results in lower cost savings regardless of the initial SOC level.