LOCAL VERSUS CENTRALIZED CHARGING STRATEGIES FOR ELECTRIC VEHICLES IN LOW VOLTAGE DISTRIBUTION SYSTEMS

Transcription

1 LOCAL VERSUS CENTRALIZED CHARGING STRATEGIES FOR ELECTRIC VEHICLES IN LOW VOLTAGE DISTRIBUTION SYSTEMS Presented by: Amit Kumar Tamang, PhD Student Smart Grid Research Group-BBCR 1 Supervisor : Prof. Weihua Zhuang 23 August, 2012

2 MAIN REFERENCE Richardson, P.; Flynn, D.; Keane, A.;, "Local Versus Centralized Charging Strategies for Electric Vehicles in Low Voltage Distribution Systems," Smart Grid, IEEE Transactions on, vol.3, no.2, pp , June

3 OUTLINE Introduction (2) Methodology (5) Simulation Model (4) Result and Discussion (6) Conclusion (1) 3

4 INTRODUCTION (1/2) Smart Grid also Envisioned to deal with future largely deployed EVs issues. Such as efficient charging of EVs (Capacity, Cost), V2G (Energy Storage) etc. Charging Strategies : Uncontrolled vs Controlled Controlled : accommodates Large number of Vehicles in distribution systems Controlled Charging: Local control vs Central Control Low Voltage Distribution system: Voltage Range from 120 V to 600 V; 3 phase to single phase 4

5 INTRODUCTION (2/2) Optimizing charging rates of EVs based on Local control charging (LCC) method. Deliver maximum amount of energy to the EVs maintaining the network in acceptable limits. Rate : determined based on local network conditions and battery state of charge. Investigates LCC in terms of network capacity utilization and total energy delivered to EVs and compared with Centralized Control Charging (CCC) method. 5

6 METHODOLOGY (1/5) Linear Programming tool in Matlab & Optimization Assumptions: EV charging units: load control capability ( can charge from zero to maximum rated charge) is in each household with an EV present. AMI, time-of-day electricity tariffs. Only grid to vehicle considered. For CCC method: DSO (Distribution system operator) controls EV charging unit from remote. Each EV connection via standard single-phase line. 6

7 METHODOLOGY (2/5) LCC Strategy: Objective: maximize amount of power delivered in their EV at each 15 min time step. Subjected to instantaneous voltage at customer point of connection (CPOC) and loading of its single phase service cable. Predetermined sensitivities of CPOC voltage and service cable loading. And No updates. Charging unit can t communicate with other charger unit on the feeder. Without Exceeding either Local voltage or single phase loading limits. 7

8 METHODOLOGY (3/5) LCC Strategy: Objective Function: P EV : Power delivered; continuous control variable: 0 kw max power o/p of charger x : 0 when EV fully charged or disconnected : 1 when EV battery not fully charged Constraints: Charging rate limits Avoid large variations of Charging rate- Δ: Defined limit, in KW, Acceptable Network Voltage limit Thermal Loading Limit of Service Cable: Lsc :Total current flowing through cable Lscmax: CPOC fuse current rating 8

9 METHODOLOGY (4/5) LCC Strategy Network Sensitivities: Predetermined Network Voltage & Loading Sensitivities with addition of EV load AND used for all the time. Don t match with continuously varying load on feeder. Determined by series of unbalanced, 3-phase load flow calculations on test network, Simulation based. Residential Base Load (Model) during charging = 2 kw. μ (V/kW) : Voltage Sensitivity β (A/kW) : Current Sensitivity Vinit : Initial voltage at CPOC 9

10 METHODOLOGY (5/5) CCC Strategy Involves voltage at each CPOC, thermal load at each single phase line to household, LV transformer, 3-phase main cable supplying feeder, Battery State of charge (BSOC) for each connected EV. Sensitivities calculated at each time step. More insight into the network, best use of network capacity o o o Objective Function: N : Number of Customer being served. BSOCi & BSOCmax : current BSOC and maximum battery capacity. Weighted objective function as per current BSOC, giving priority to EVs with low BSOC, even distribution of energy. Assumption: Necessary monitoring and communication equipment is installed on feeder. 10

11 SIMULATION MODEL (1/4) Test Network: o LV residential distribution feeder in suburban area of Dublin, Ireland. 74 households. 3 phase main cable 432 m 1 phase Service cable 2.16 km Nominal voltage of 230/400V tolerance 10% Specification by Electricity Supply Board (ESB), DSO of Republic of Ireland 11

12 SIMULATION MODEL (2/4) Load Modeling 15 min time series demand data from DSO for high, medium and low use customers, Randomly assigned houses in test network. Off-peak tariff : 11 p.m to 8 a.m (following day) Residential load power factor = 0.95 inductive, 50% constant power and 50% constant impedance. EV battery capacity = 20 kwh Max Charge Rate = 4 kw (90% efficient) EV load is power load with unity power factor. 12

13 SIMULATION MODEL (3/4) Time Period for Investigation:12 noon to 12 noon (following day) in January (Winter); Sample 24-h time period from data. EVs Penetration: 50% (37 out of 74 households) EVs demand (max) : 37x 4 kw = 148 kw Random Location (Fixed), connection time (+/- 3 hrs of 11 p.m) and connection duration (6 to 15 hours) Initial BSOC assigned = 0 % to 75% of 20 kwh. 13

14 SIMULATION MODEL (4/4) Stochastic Scenarios: In contrast to specific network scenario, generate different residential load scenarios. Probabilities Distribution Functions (PDFs) of household load created based on data by DSO: and 15 min household load profiles generated from 12 noon to 12 noon (following day). Varying residential load, EV location, initial BSOC and duration of connection at each 24-h period. Rest are similar with pervious case. 14

15 RESULT AND DISCUSSION (1/6) Uncontrolled EV charging: EV once connected, charged at maximum rate of 4 kw until it reaches full BSOC. Maximum number of EVs allowed in test network with uncontrolled scenario = 7 (10% of households) 15

16 RESULT AND DISCUSSION (2/6) Controlled EV charging: For optimization process: lower voltage limit 0.92 pu contrast to 0.9 pu for safety margin. - Δ in (3) is set at 1 kw In (6) μ = to V/kW. ( CPOC at extremes of feeder more sensitive.) β = 8.2 to 8.7 A/kW (calculated) Crossing 0.92 pu Margin. Shifting of Demand Later at night 16

17 RESULT AND DISCUSSION (3/6) Maintain 0.92 pu Voltage label Much tighter Demand Profile Required 517 kwh demand achieved in less time No charging 17

18 RESULT AND DISCUSSION (4/6) LCC: 3 out of 37 EVs has less that 100% charge in contrast to all EVs in CCC being fully charged. CCC utilizes Network Capacity more 18

19 RESULT AND DISCUSSION (5/6) Loading service cable Is not binding constraint For optimization. (Max is just below 80%) In contrast to lowest CPOC voltage. 19

20 RESULT AND DISCUSSION (6/6) N/w utilization: More occurrence at lowest voltage For CCC Majority loading Below 60%. Thermal loading is not Binding constraints. CCC: 100% of final BSOC found to be in 95%-100% unlike LCC Stochastic Scenarios Analysis: test for 300 distinct 24-h periods (28800 time steps) during winter. instead of (24x4=96 time steps) 20

21 CONCLUSION Controlled charging brings high penetration of EVs on LV network as compared to Uncontrolled Charging. LCC may require larger safety margins to maintain network parameters as compared to CCC. CCC has better network utilization. LCC implementation requires less communication infrastructure. With AMI, LCC may be sufficient for initial penetration. 21