Analysis of electric vehicle charge scheduling and effects on electricity demand costs

Due to the recent inflow of electric vehicles (EVs) to the automobile market, new concerns have risen with respect to the additional electrical load and the resultant effects on an overloaded electric grid. It takes approximately three to four hours to fully charge the battery of some common EVs depending on battery capacities. Either for convenience purposes or possibly necessity due to limited electric range on EVs, some EV owners may desire to charge their EV while at work in addition to charging at home. These forward-thinking daytime charging providers are typically Commercial and Industrial (C&I) electric ratepayers, or other large electric consumers which constitute the majority of businesses, shopping centers, academic campuses and manufacturing facilities. C&I consumers generally pay for electricity consumption based not only on total energy consumed, but also on peak electric demand measurements because certain portions of the working day are traditionally associated with C&I peak load periods. Electricity consumption during these periods results in higher electricity costs due to differences in the billing structures. Therefore, uncontrolled EV charging in large numbers potentially creates significant peak demand increases resulting in higher electric costs for the C&I ratepayer and exacerbating grid issues relating to electric generation and grid stability. Proper scheduling of EV charging is a necessity to avoid such increases in peak electric demand. Scheduling can be implemented in a number of ways including traditional process control strategies such as model predictive control or machine scheduling algorithms. When viewed as a parallel machine scheduling problem, charging stations can be considered as identical machines and the EV charging transactions required can be seen as jobs with varying processing times and due dates. The goal of such an algorithm is to schedule EV charging so that total electricity cost is minimized but off-peak charging activity is maximized. Due to the nature of electricity billing structures and cost calculations for C&I ratepayers, the machine scheduling algorithm must use time index based operations. This paper examines the EV charge scheduling problem and casts it as a parallel machine scheduling problem with availability constraints. Particularly, this study develops several heuristic scheduling algorithms such as Greedy Local Search and Simulated Annealing. These algorithms are evaluated using MATLAB (Mathworks MATLAB R2012b, 2012) to compare performance. In addition, the general-purpose optimization solver CPLEX (Ibm, ILOG CPLEX, 2012) is used to determine exact solutions for small size problems, which is then used to benchmark the performance of the other algorithms studied.