(c) 2016 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2016.2647620, IEEE Transactions on Smart Grid TSG-01131-2016.R1 1 Real-Time Smart Charging of Electric Vehicles for Demand Charge Reduction at Non-Residential Sites Guanchen Zhang, Student Member, IEEE, Shaoqing Tim Tan, G. Gary Wang Abstract—Smart Electric Vehicle (EV) charging deals with increasing demand charges caused by EV load on Electric Vehicle Supply Equipment (EVSE) hosts. This paper proposes a real- time smart charging algorithm that can be integrated with Commercial & Industrial (C&I) EVSE hosts through Building Energy Management System (BEMS) or with utility back office through the Advanced Metering Infrastructure (AMI). The proposed charging scheme implements a real-time water-filling algorithm (RTWF-n1) able to reduce the peak demand and to prioritize EV charging based on the data of plugged-in EVs. The algorithm also accommodates utility and local Demand Response and Load Control (DRLC) signals for extensive peak shaving. Real-world EV charging data from different types of venues are used to develop and evaluate the smart charging scheme for demand charge reduction at Medium & Large General Service locations. The results show that even at constrained venues such as large retails, monthly demand charges caused by EVs can be reduced by 20-35% for 30% EV penetration level without depreciating EVs’ charging demand. Index Terms—Electric Vehicle, Smart Charging, Demand Charge, Peak Shaving, Demand Response, Direct Load Control. NOMENCLATURE Abbreviations AMI Advanced Metering Infrastructure BEMS Building Energy Management System C&I Commercial & Industrial DCFC DC Fast Charger DR Demand Response DRLC Demand Response and Load Control EV Electric Vehicle EVSE EV Supply Equipment L2 Level-2 Charging (208/240V) LB Lower Bound LGS Large General Service MGS Medium General Service OPEX Operational Expense PHEV Plug-in Hybrid Electric Vehicle RTWF Real-Time Water Filling SC Smart Charging SOC State of Charge TOU Time of Use Variables Δt Smart meter sampling period, or the SC control time step Manuscript submitted on Aug 24, 2016; revised on Nov 29, 2016 The authors are with the School of Mechatronic Systems Engineer- ing, Simon Fraser University, Surrey, B.C., Canada. Emails: gza10@sfu.ca; ttan@sfu.ca; gary wang@sfu.ca. η Average charging efficiency t ∗ u Unplug time of EVs in {EV 1 } as of t m τ 1 , τ 2 Start and end time of the DRLC event θ Charging flexibility of the i ′ th EV  E Average deferrable energy for SC in an EV {EV 1 } Group of EVs with the earliest leaving time as of t m {EV (−1) } Group of EVs not in {EV 1 } as of t m a Binary term indicating whether the i ′ th EV is plugged in at time t E d Total energy demand of an EV E ′ d New feasible energy demand after curtailment E (+) Maximum energy can be charged to {EV (−1) } before all EVs in {EV 1 } leave E (−) Minimum energy must be charged to {EV (−1) } before all EVs in {EV 1 } leave E (j) ǫ Remaining energy to be charged as of t m for the j ′ th EV in {EV 1 } L 0 Original charging power profile L FD Power profile of fully deferrable charging session L PD Power profile of partially deferrable charging session N t Number of EVs plugged in at t N ν Number of unplugged EVs (done charging) at t = t u N EV Total umber of EVs come to charge in a day N ∗ EV Number of EVs in {EV 1 } p c Constant-current charging power of an EV p λ Provisional lower bound for handling uncertainties towards t s P υ Upper bound of building-level demand by DRLC at time t p υ Maximum charging power subject to SOC p ζ Applied lower bound for the charging power of the i ′ th EV at t p min Minimum charging power limited by EVSE hardware or software T Number of control steps for SC in a day T F Average flexible duration for SC t m Current time in a SC control window t p Plug-in time of an EV T s Number of control steps in each real-time SC window t s End time of a SC control window t ∗ s Earliest unplug time of EVs in {EV 1 } t u Unplug time of an EV x (i) Charging power of the i ′ th EV at t ∈ [t m ,t s ] x b Building’s non-EV demand at time t X EV Total demand of all plugged-in EVs at time t Y Aggregate load profile including building and all EVs y t Total demand in kW at time t