- Email: [email protected]

Contents lists available at ScienceDirect

Energy Conversion and Management journal homepage: www.elsevier.com/locate/enconman

Strategic Energy Management (SEM) in a micro grid with modern grid interactive electric vehicle Lokesh Kumar Panwar a, K. Srikanth Reddy b, Rajesh Kumar a,⇑, B.K. Panigrahi b, Shashank Vyas a a b

Center for Energy and Environment, MNIT Jaipur, India Indian Institute of Technology, Delhi, India

a r t i c l e

i n f o

Article history: Received 23 June 2015 Accepted 8 September 2015

Keywords: Micro grid Electric vehicle (EV) Unitized regenerative fuel cell (URFC) Particle swarm optimization (PSO) Forward/backward sweep algorithm Heuristic approach

a b s t r a c t In this paper, strategic energy management in a micro grid is proposed incorporating two types of storage elements viz. unitised regenerative fuel cell (URFC) and electric vehicle (EV). Rather than a simple approach of optimizing micro grid operation to minimize line loss in the micro grid, this paper deals with multi objective optimization to minimize line loss, operational cost and maximize the value of stored energy of URFC and EV simultaneously. Apart from URFC, two operation strategies are proposed for EV enabling V2G operation to reduce overall system cost of operation. To address the complexity, non-linearity and multi dimensionality of the objective function, particle swarm optimization-a heuristic approach based solution methodology along with forward and back sweep algorithm based load flow solution technique is developed. Combined with particle swarm optimization (PSO), forward and backward sweep algorithm resolves the complexity and multi dimensionality of the load flow analysis and optimizes the operational cost of micro grid. The simulation results are presented and discussed which are promising with regard to reduction in line loss as well as cost of operation. Scheduling strategy of the micro grid with both URFC and EV enabling V2G operation presents a promising approach to minimize line loss and cost of operation. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Alarming environmental concerns glooming over climate change and global warming have emphasized the use of sustainable practices to be a potential means of emission curtailment from sectors like power generation and transportation [1]. Renewable energy technologies (RET) and green transportation through electric and hybrid vehicles are identified as potential contenders in such emission curtailment initiatives [2]. While RETs deployment has been in lime light from past decade facilitated by various fiscal policies all over the world [3]. Lack of suitable and cost effective on board storage technologies hampered effective EV deployment. However, electrification of transport through EV and hybrid vehicle looks imperative as 13% of global greenhouse gas (GHG) emissions were due to emissions from fossil fuel combustion in conventional vehicles [4]. The EV deployment as a greener mode of transportation is also supported by fiscal policies like initial financial assistance and incentivized charging rates. This is conceived because of interests of nations like US and UK in promoting electrification ⇑ Corresponding author. E-mail addresses: [email protected] (L.K. Panwar), srikanthkonda. [email protected] (K.S. Reddy), [email protected] (R. Kumar), bijayaketan. [email protected] (B.K. Panigrahi), [email protected] (S. Vyas). http://dx.doi.org/10.1016/j.enconman.2015.09.019 0196-8904/Ó 2015 Elsevier Ltd. All rights reserved.

of transportation. The U.S has put an ambitious goal of manufacturing 1.2 million EVs and bringing 1 million EVs on road by 2015 to reduce aggravated dependence on fossil fuels like oil and gas [5]. Similarly, in the UK also more than 80% of car owners intend to replace their conventional vehicle with electric vehicle and about 73% of them would consider buying EV by 2015 [6,7]. The main motivation for this being cost savings through electrified transport and facilities like parking, incentivized charging, tax benefits over usage of EV. Therefore, in near future the number of EVs and their effect on existing electric network would be pronounced if not managed properly. Initial research was focused on analyzing the consequences of EV charging loads on electric utility and system security constraints like frequency, voltage regulation issues [8]. Thereafter, many studies carried out management of charging load through optimizing the EV charging tasks [9–11]. Later, vehicle to grid operation was found to be an option to avail the on board storage of EV or hybrid vehicle for utility needs [12]. The cost and revenue perspectives of EV in V2G operation for possible potential applications like energy source, spinning reserve and regulation services were analyzed [13]. Such market participation of EV offering V2G services presents an encouraging revenue option for EV owner in extending financial viability of EV adoption. Later, V2G operation

42

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

Nomenclature Ii;j ðtÞ; Imax i;j current flow between i, j buses during time t and maximum current carrying capacity of line Pload ðtÞ; Q load ðtÞ active and reactive power loads of system for tth hour V i;min ; V i;max maximum and minimum voltage limits for ith bus Ppv ðtÞ; Eev ðtÞ power generated by PV and energy of EV at time t SOC ev ;min ; SOC ev ;max minimum and maximum SOC levels of EV Pev ;max ; Eev ;max maximum power and energy limits of EV V ev ; V URFC voltage levels of EV and URFC storage elements PURFC ðtÞ; EURFC ðtÞ power and energy levels of URFC at time t Q 1;min ; Q 1;max reactive power limits of PCC (bus/node 1) P1;min ; P1;max active power limits of PCC (bus/node 1) Pi ; Q i active, reactive powers of ith bus

is extended to intelligent scheduling of EV along with thermal generators to reduce both cost and emissions [14]. Apart from thermal units, credibility of V2G operation is also checked for systems with renewable energy technologies (RET) penetration [15]. Meanwhile, the integration of EV is also considered for distribution networks to alleviate its effectiveness in offering energy and ancillary services [16]. The micro grid technology comprises a key feature of future distribution system in rapidly evolving electric utility networks. In such micro grid networks, distributed energy resources like renewable energy and distributed energy storage (DES) elements play an important role in energy management [17]. Often, solar photovoltaic and storage systems like batteries and fuel cell systems comprise distributed energy resources and distributed storage devices in a micro grid [18]. In terms of long term use, fuel cell systems has an edge over batteries with shorter life time [19]. The energy management in such systems is carried out through minimization of operational cost and line loss. Effective management strategies are required for such cost minimal operation and optimizing charge and discharge scheduling of storage devices is imperative. In addition to the available storage systems like battery bank systems, fuel cell systems, utilization of on board storage of EV can deliver valuable returns for both utility and EV owner in a micro grid structure. Also, the uncertainty of EV movement can be handled effectively in a micro grid compared to large scale EV deployment through aggregation policies. In literature, energy management and optimal power flow (OPF) problem in micro grid is addressed by many methodologies like dynamic programming, linear programing, non-linear programming, interior point and newton based methods [20]. In general, gradient based and sensitivity based approaches are used to obtain OPF through linearized objective function associated with networks constraints. However, the traditional optimization approaches are limited in application for objectives with nonlinear and complex constraint attributes and may get struck in local minima rather than reaching a global minimum [21]. This problem is addressed in this work through integration of forward/backward sweep algorithm for power flow with heuristic approach like particle swarm optimization (PSO) [22–25] to solve OPF for globally optimal solution. In earlier research, PSO is used to model the storage elements like Li-ion batteries, which forms a pivotal component in micro grid structure with renewable energy sources [26]. Also, PSO is deployed for state of charge estimation in storage applications [27]. This paper discusses a strategic management of distributed storage elements namely EV storage and fuel cell in micro grid to avail economic as well as environmental benefits. The EV is scheduled in conjunction with distributed energy resources

V i; V j Ij R; X; Z

voltages at bus i and j respectively current flowing from bus j line resistance, reactance and impedance between bus i and j gE ; gFC fuel cell efficiencies in electrolysis (E) and fuel cell (FC) mode respectively PL ðtÞ; CðtÞ line loss and electricity price at tth hour gch ; gdis charging and discharging efficiencies of EV battery Pev ði; tÞ power dispatched to/consumed from grid by ith EV at tth hour Uði; tÞ availability of ith EV at tth hour E24; Eev ði; 24Þ residual energy of URFC and ith EV at 24th hour respectively

(DER) and fuel cell system to minimize operational cost. The possible energy management strategies emphasizing intelligent charging schedule and V2G operation are developed. The proposed strategies are implemented in micro grid considering operational characteristics of fuel cell and EV storage along with consideration of EV mobility patterns, state of charge with respect to time of operation and load. Rest of the paper is organized as follows: network description with solar photovoltaic (SPV), EV and URFC is introduced in Section 2. Thereafter, problem formulation with energy management strategies for given load, generation and storage elements is developed in Section 3. Solution methodology for energy management using heuristic approach is outlined in Section 4. Simulation results of the proposed methodology for different strategies are produced and discussed in Section 5. Finally, Section 6 concludes the work with key findings and future scope. 2. System description In this work, a typical micro grid structure with conventional generator along with SPV for generation is considered and two EVs, URFC are considered as storage elements. The single line diagram of network (Fig. 1) incorporates per unit line impedances for a base voltage of 13.8 kV and a base power 5 MVA. Load demand and generation of SPV for capacities of 1 and 0.5 MWp are taken from [24] and are shown in Fig. 2. The hourly load considered is distributed amongst available loads randomly for a time horizon of 24 h. For the network considered, power balance and voltage profile constraints are checked through performing load flow analysis. Conventional techniques like newton raphson method and fast decoupled methods are not used for carrying out load flow analysis considering their incompatibility to solve the ill conditioned radial distribution systems [25]. Therefore, features like computational simplicity and efficiency of forward/backward sweep algorithms are exploited to solve load flow problem of distribution network with radial nature. Thus, a modified forward/ backward sweep algorithm is used in this work in conjunction with EV and URFC. The objective function is associated with some network constraints described as follows. 2.1. Constraints 2.1.1. Power line constraint At any time, current flowing through a line between two buses should be within the maximum current carrying capacity between two buses.

Ii;j ðtÞ < Imax i;j

ð1Þ

43

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

Fig. 1. Micro grid line diagram with DER and DES [25].

(a) kW

5,000 2,500 0

2

4

6

8

10

12

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

(b) kVAr

2000 1000 0 -1000

2

4

6

8

10

12

(c) kW

1000 500 0

2

4

6

8

10

12

(d) kW

500 250 0 2

4

6

8

10

12

time (hour) Fig. 2. (a) Active power of load demand, (b) reactive power of load demand, (c) power output of 1 MW PV plant and (d) power output of 0.5 MW PV plant.

44

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

2.1.2. Load constraints The distributed loads in the given network would be operating under constraints like active, reactive power in given voltage limits as given by,

P load ðtÞ ¼ þPload ðtÞðfixedÞ Q load ðtÞ ¼ þQ load ðtÞðfixedÞ

ð2Þ

V i;min 6 V i ðtÞ 6 V i;max

equation for receiving end voltage as a function of sending end voltage and power can be deduced by re arranging Eqs. (7), (8) and is given by,

sﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ðP2 Q 2 Þ V j ¼ V 2i 2ðPi R Q i XÞ þ i 2 i Z 2 Vi

ð9Þ

V i ¼ V j þ Ii Z

ð10Þ

V j ¼ V i Ii Z

ð11Þ

2.1.3. SPV plant constraints Like the conventional generation source, the SPV plant has to operate under the following equality and inequality constraints.

where V i and V j are the sending end and receiving end voltages calculated during the load flow analysis of radial networks through forward/backward sweep algorithm.

P PV ðtÞ ¼ PPV ðtÞðfixedÞ

2.2. URFC system

Q load ðtÞ ¼ Q PV ðtÞðfixedÞ

ð3Þ

V i;min 6 V i ðtÞ 6 V i;max 2.1.4. EV battery constraints In this study, LiFePO4 battery is considered to operate under charging and V2G modes. At any given time, the battery storage unit of EV has to operate under constraints of state of charge (SOC limits), power limits as shown below.

V i ðtÞ ¼ V ev SOC ev ;min 6 SOC ev ðtÞ 6 SOC ev ;max 0 6 Eev ðtÞ 6 Eev ;max 0 6 Pev 6 Pev ;max

ð4Þ

Based on the superior cycle life and discharge characteristics of LiFePO4, lower and upper limits of SOC are considered as 0, 1 respectively. 2.1.5. URFC constraints The fuel cell system considered in this work would be operating in two modes, fuel cell (FC) mode and electrolysis (E) mode under the following constraints.

In this work URFC is considered as a distributed storage element serving as a storage medium for hydrogen and electrical energy. In the micro grid network, a 30 kW unit of URFC is considered with following operation.

1 H2 þ O2 $ H2 O þ DH 2

ð12Þ

where DH denotes the enthalpy of chemical reaction (12) which is equal to ±237.2 kJ/mol. The amount of hydrogen consumed in electricity generation or produced from electricity is given by,

gsys PURFC 3600 moles ¼ H2 in h DH

ð13Þ

where

gsys ¼

gE ¼ 90% if P URFC is positiv e gFC ¼ 50% if P URFC is negativ e

ð14Þ

where in equation (14), gE and gFC denote the efficiencies of electrolysis mode and fuel cell modes respectively. The maximum efficiency of 93% and 53% are achieved in E mode and FC respectively [23].

V i ðtÞ ¼ V URFC 0 6 EURFC ðtÞ 6 EURFC;max In FC mode PURFC;max < Pi ðtÞ < PURFC;min

2.3. EV system

ð5Þ

In E mode þ PURFC;min < Pi ðtÞ < þPURFC;max 2.1.6. Point of common coupling (PCC) constraints The micro grid and associated components are connected to utility grid through point of common coupling (point 1 in Fig. 1) which is maintained under following constraints.

d1 ðtÞ ¼ 0 V 1 ðtÞ ¼ V in ðfixedÞ Q 1;min 6 Q 1 6 Q 1;max

ð6Þ

P 1;min 6 P1 6 P1;max The load flow equations at the receiving end are given by,

V 2j ðV i V j Þ Pj ¼ cosðhz di þ dj Þ cosðhz Þ Z Z 2 Vj ðV i V j Þ sinðhz di þ dj Þ sinðhz Þ Qj ¼ Z Z

Along with URFC, pure electric vehicle/EV equipped with a 300 kW h Lithium iron phosphate battery is also considered in this work [29]. The LiFePO4 battery has a long life of 3500 cycles and a capacity retention of 80% at 100% depth of discharge (DOD) [30]. Therefore, it presents a full deep discharge operation at good cycle life unlike the standard Li-ion battery with a DOD limit of 80% upon which voltage profile and cycle life are adversely effected.

0 6 SOC ev 6 1

ð15Þ

The EV considered is operated in charging as well as V2G operation with an efficiency of 90%. In this work EV is used for transportation of employees inside the industrial micro grid. Each EV is scheduled for 3 trips daily with 40 km per trip. Therefore, during one trip time a constant energy consumption of 50 kW h occurs approximately. 3. Problem formulation

ð7Þ ð8Þ

In Eq. (7), Vi, Vj denote the voltages at sending and receiving ends respectively, Z stands for per unit impedance, hz, di, dj are phase angles of impedance and voltages at both ends of the line. The

The operation and load flow of micro grid is governed by line losses. The line loss can be determined using incoming power at the point of common coupling. The line losses and cost of electricity in the micro grid will determine the operational strategy of EV and URFC in micro grid structure. This work considers forward/ backward sweep algorithm to determine load flows in which voltages at nodes are evaluated by forward sweep and line currents

45

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

with associated power losses are evaluated through backward sweep. The following line loss formula is used to evaluate the monetary value of line losses.

F1 ¼

24 X PL ðtÞ CðtÞ

ð16Þ

t¼1

where P L ðtÞ and CðtÞ denote the total line losses and price of electricity at tth hour respectively. In the micro grid structure URFC and EV are used at buses 6 and 5 respectively. Three operation strategies are proposed to evaluate the effectiveness of URFC and EV in micro grid operation. 3.1. Case I: Micro grid operation strategy with UFRC In this case, URFC is used at bus 6 to reduce the operational cost and for reliability improvement of the micro grid. While, conventional vehicle is used for transportation in the micro grid. The cost of operation for URFC can be calculated using the following equation.

F2 ¼

24 X PURFC ðtÞ CðtÞ

ð17Þ

t¼1

where

PURFC ¼

> 0 for charging

ð18Þ

< 0 for discharging

In Eq. (18), PURFC denotes the power drawn from grid or delivered to grid by the URFC system and C(t) is the price of electricity for tth hour. The amount of energy (kW h) that, URFC is left with, in the penultimate day is used in the next day and cost of the energy left is given by,

F 3 ¼ E24 C Av g geff

ð19Þ

where E24 is the energy left with URFC in 24th hour of the day, CAvg is the average tariff of the day and geff is the overall system efficiency. The overall objective is of dual nature associated with minimization of F1, F2 and maximization of F3. Therefore, main objective is minimizing line losses and cost while maximizing the URFC residual energy for the day as given by,

Obj:case

I

( ) 24 24 X X ¼ Min PL ðtÞ CðtÞ þ PURFC ðtÞ CðtÞ E24 C Av g geff t¼1

t¼1

ð20Þ The energy of fuel cell at tth hour is given by,

(

EURFC ðtÞ ¼

EURFC ðt 1Þ þ gch PURFC ðtÞ if P URFC ðtÞ > 0 ðtÞ EURFC ðt 1Þ þ PURFC g

if P URFC ðtÞ < 0

dis

ð21Þ

3.2. Case 2: Micro grid operation strategy with URFC and EV charging (without V2G) The operational cost and lines losses can be reduced further by deploying EV for off peak charging. In this case, micro grid is operated with EV and URFC at buses 5, 6 respectively. The EV is charged at cheap hours and used for transit in micro grid.

Obj:caseII

( 24 24 X X ¼ Min: PL ðtÞ CðtÞ þ PURFC ðtÞ CðtÞ t¼1

t¼1

) N X 24 X Uði; tÞPev ði; tÞ CðtÞ E24 C Av g geff ; þ i¼1 t¼1

Uði; tÞ 2 f0; 1g8i; t

ð22Þ

where N is the number of electric vehicles, geff is the system overall efficiency, U(i, t) is a binary variable (0, 1) that determines whether EV is in transit or not. While the EV is in transit, U(i, t) is 0 and vice versa.

Uði; tÞ ¼

0

if t 2 T tr ðiÞ

ð23Þ

1 otherwise

where Ttr(i) denotes the time for which ith EV remains in transit. 3.3. Case 3: Micro grid operation with URFC and V2G operation The V2G concept has proven to be cost effective when deployed in smart grid operation. Therefore, in this case such V2G operation is devised taking the energy consumed, delivered between grid and in transit into account as given by,

8 > < Eev ði; t 1Þ þ gch P ev Uði; tÞ if Pev > 0 and t R T tr;i P ev Uði; tÞ if Pev < 0 and t R T tr;i Eev ði; tÞ ¼ Eev ði; t 1Þ þ gdis > : Eev ði; t 1Þ Eev ;trip if t 2 T tr;i ð24Þ where Eev(i, t) is the energy of ith EV at tth hour, Eev,trip is the battery energy consumed by EV in transit mode and energy exchanged with grid (Pev) is zero. The objective with V2G operation is to minimize EV and URFC charging cost, line loss while maximizing the energy at end of the day i.e., 24th hour. The overall objective function in case III is given by

Obj:case

( 24 X

III

¼ Min

PL ðtÞ CðtÞ þ

t¼1

24 X PURFC ðtÞ CðtÞ t¼1

N X 24 X Uði; tÞPev ði; tÞ CðtÞ E24 C Av g geff þ i¼1 t¼1

) N X Eev ði; 24ÞC Av g geff ; Uði; tÞ 2 f0; 1g8i; t

ð25Þ

i¼1

where Eev(i, 24) is the residual energy of ith EV after three operating modes charging, V2G and transit modes for the day. 4. Solution methodology The operation of micro grid is divided into three cases where one case does not include EV and remaining two cases deploy EV (Flow chart 1). Since, when the vehicle is in transit, grid interaction is not possible the strategy using URFC is employed for that particular hour. However, it is not obligatory for the vehicle to work in either case II or case III, rather the vehicle can be idle if the tariff offered by grid is not profitable or suitable. The energy of URFC and EV are calculated using efficiency in that particular mode of operation and the tariff offered. The charging and V2G operation in this work is modelled for individual EV but not aggregated. Therefore, scheduling of each EV operation is independent of each other. The use of URFC and EV in this work to reduce line loss and cost of operation when operated in conjunction with PV system is an optimization problem. However, with the integration of storage elements like URFC, EV and their operating constraints, the OPF problem gets to be intricate, nonlinear and hard to tackle. The intricacy lies in the way that we are scheduling the operation for the whole day, i.e. for 24 h rather than for single hour. The use of heuristics is not imperative in case of URFC, EV scheduling for a single hour through minimization of line loss and cost of operation. On the other hand, optimizing the operation schedule for the whole day with variable energy price every hour makes the problem multidimensional and furthermore, complex in a manner in

46

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

(a) INR/kWh

30 20 10 0

2

4

6

8

10

12

14

16

18

20

22

24

14

16

18

20

22

24

15

17

19

21

23

25

14

16

18

20

22

24

(b)

kW

25 0 -25 2

4

6

8

10

12

(c)

kWh

100 50 0 3

5

7

9

11

13

(d)

MW

5 4 2 0 2

4

6

8

10

12

time (hour) Fig. 3. (a) Differential tariff, (b) power drawn/supplied by URFC, (c) energy contained by URFC and (d) power at PCC.

light of the fact that planning URFC operation every hour requires to know its impact on line losses and stored energy at preceding hours. Simultaneous scheduling of both storage elements URFC and EV with their respective constraints and network constraints makes the problem complex. Also, the variable cost of electricity and EV transit in the network adds dynamism to the problem. To solve multiple objectives simultaneously, PSO is deployed for scheduling EV and URFC. PSO mimics the social behavior and communication strategies of flock individuals in searching most promising food source. The search approach follows updating both velocity and direction based on particle’s own and group’s experiences. This is enabled by tracking the best possible solution/position known so far (Pbest) and the best solution known to neighboring individuals of the group (Gbest). Every particle updates its velocity and trajectory of direction based on following velocity and direction updates.

V tþ1 id

¼ w:V tid þ C 1 :rnd1 :ðPtbest xitþ1 ¼ V tþ1 þ xti i

xtid Þ

þ

C 2 :rnd2 :ðGtbest

and backward sweep algorithm consists of three steps as given below [24]. 1. Calculation of node currents: All the nodes (buses) are sorted staring from source/root node and farthest node is designated as nth bus. For kth iteration, nodal current of ith node is given by,

Iki ¼

Si

!

V k1 i

ð26Þ

Firstly to begin the search, Np number of particles, each of dimension T N, are created in the search space. Here ‘‘T” stands for the number of hours for which the scheduling is done i.e. 24 h and N is the number of storage elements which is 3 (1 URFC and 2 EVs). Each and every particle is then tested for fitness after putting it in the objective function and evaluating the output keeping in view the problem constraints. Subsequently the appropriately fit solutions are retained and every particle is updated in the process. Then the forward–backward sweep algorithm is applied to calculate the micro grid losses considering the network topology [28]. The voltage profile is assumed to be flat given the root/source node. The forward

ð27Þ

where Si represents power injected at node i, V k1 is the volti age ith node voltage calculated in k 1th iteration and Yi denote shunt elements sum at ith node. 2. Backward sweep: Starting at farthest node (i), for every iteration currents in other branches moving toward root node are calculated using KCL as:

Iki1;i ¼ Iki þ

xtid Þ

8i ¼ 1; 2; . . . n

Y i V k1 i

X

currents emanaing from node i

8i ¼ n; n 1; . . . 1

ð28Þ

where Iki is the current injected at node i. 3. Forward sweep: Forward sweep starting form root node to farthest node is run to calculate nodal voltages. Through KVL and using current (calculated in step 2) flowing from i 1th node to ith node i.e., Iki1;i the nodal voltages can be calculated as:

V ki ¼ V ki1 Z i1;i Iki1;i

8i ¼ 1; 2; . . . n

ð29Þ th

where Zi1,i is the line impedance between ði 1Þ and ith nodes. Steps 1 to 3 are repeated until convergence criteria is satisfied.

47

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

INR/kWh

(a) 30 20 10 0

2

4

6

8

10

12

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

15

17

19

21

23

25

15

17

19

21

23

25

15

17

19

21

23

25

14

16

18

20

22

24

(b) kW

25 0 -25 2

4

6

8

10

12

(c) kW

100 50 0

2

4

6

8

10

12

(d) kW

100 50 0

2

4

6

8

10

12

(e) kWh

100 50 0

3

5

7

9

11

13

(f) kWh

200 100 0

3

5

7

9

11

13

(g) kWh

200 100 0 3

5

7

9

11

13

(h) MW

5 4 2 0

2

4

6

8

10

12

time (hour) Fig. 4. (a) Differential tariff, (b) power drawn/supplied by URFC, (c) charging status of EV1, (d) charging status of EV2, (e) energy contained by URFC, (f) energy status of EV1, (g) energy status of EV2 and (h) power at PCC.

Once the nodal voltages, currents are obtained using forward– backward sweep algorithm, losses are calculated. The calculated losses are then multiplied with the tariff to find the cost which is added to the current fitness value to get the total fitness value.

Update process is then applied to the PSO to find new particles. Based on the quality of solutions and the stopping criteria of the optimization process, the complete process keeps on repeated iteratively or terminated.

48

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

(a) INR/kWh

30 20 10 0

2

4

6

8

10

12

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

14

16

18

20

22

24

15

17

19

21

23

25

15

17

19

21

23

25

15

17

19

21

23

25

14

16

18

20

22

24

(b) kW

25 0 -25 2

4

6

8

10

12

(c)

kW

100 0 -100

2

4

6

8

10

12

(d) kW

100 0 -100

2

4

6

8

10

12

(e) kWh

100 50 0

3

5

7

9

11

13

(f) kWh

400 200 0

3

5

7

9

11

13

(g)

kWh

400 200 0

3

5

7

9

11

13

(h) MW

5 4 2 0

2

4

6

8

10

12

time (hour) Fig. 5. (a) Differential tariff, (b) power drawn/supplied by URFC, (c) charging status of EV1, (d) charging status of EV2, (e) energy contained by URFC, (f) energy status of EV1, (g) energy status of EV2 and (h) power at PCC.

49

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

5. Simulation results and discussion

In this case, conventional vehicles are used for transit inside micro grid and URFC is used as storage element of micro grid to reduce line loss and operational cost. The operation of micro grid resulted in line losses accounting for INR 10635 for the time horizon considered. Whereas, operational cost has reduced to INR 10496.11 with URFC integration to curtail the operational cost. It can be observed that, differential tariff is low/cheap at night and morning hours (23rd hour to 8th hour) and attained peak cost of INR 25/kW h. The charging and discharging profile of URFC system (Fig. 3(a)) depicts the off peak (w.r.t tariff) charging and discharging in hours with costly tariff. Also, the variation in charging power of URFC for same tariff rate can be attributed to the multi objective nature of system i.e., line loss minimization has prompted the change in charging power of URFC though tariff remained the same between 4th and 5th hours. The URFC charging profile shows a linear increase of stored energy (hydrogen) in the lower tariff and attains a saturated state once the capacity is filled (Fig. 3(c)). Also, the power exchanged between utility and micro grid is also in accordance with maximum allowable power of coupling transfer which can be observed through power value at PCC (Fig. 3(d)). 5.2. Case II: Micro grid operation strategy with UFRC and EV charging (without V2G) In this case, the conventional vehicles used for transit are replaced by electric vehicles/EV to avail environmental benefits in micro grid structure. The EVs are charged in off peak hours with cheap electricity prices to keep the extra energy cost to as minimum as possible. The overall objective function value of case II is INR 13860.91962 which is higher compared to case I which can be attributed to the increased energy consumption and line loss due to charging loads EVs in micro gird. Though the operation cost of micro grid is elevated due to charging loads, overall system economics have to be observed to evaluate effectiveness of EV. Considering distance travelled by both EVs during transits i.e., 240 km, comparing with conventional diesel buses with a fuel efficiency of 3 km/L the transportation economy results in INR 4128 ((240/3) ⁄ 51.6). Therefore, EV deployment for transit in micro grid results in net saving of INR 763.19 (10496.11 + 4128–13860.9) per

Line Losses (W)

40 30 20 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour Fig. 6. Micro grid line losses for different cases.

Average System Voltage (kV)

5.1. Case I: Micro grid operation strategy with UFRC

Case 1 Case 2 Case 3

50

0

13.85 13.8

Case 1 Case 2 Case 3

13.75 13.7 13.65 13.6 13.55

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour Fig. 7. Variation in average system voltage for different operation strategies.

13.82

EV bus voltage (kV)

The operation strategies of micro grid with URFC and EV as storage elements to reduce line loss and improve the cost effectiveness is simulated using PSO with 30 particles and parameter settings of 0.4, 2, 2 for w, C1 and C2 respectively. The simulation is run using MATLAB platform on a 32-bit Pentium dual core 2.3 GHz processor working on Linux operating system. URFC and EV are scheduled for a 24 h time horizon and for the load and PV generation shown in Fig. 2. The load and generation data is taken from [25]. For the time horizon, the micro grid faces a peak active load of 5 MW and peak reactive load of 1 MVAr. While PV generation from both 1 MW and 0.5 MW plants attains a peak output around 12th hour of the day. The two EV’s are scheduled for grid interaction when the particular EV is available. Transit times of EVs are modelled for typical micro grid operation. The first EV, EV1 has 3 transits for a day at 3rd, 11th and 19th hours with each one consuming 50 kW h of energy. The second EV, EV2 also undergoes three transits per day at hours 5th, 13th and 21st hours respectively with same transit energy consumption as EV1. EV efficiency in both charging and V2G operation is considered to be 90%. Three operational strategies are proposed, simulation results are presented and discussed as follows.

60

13.77

Case 1 Case 2 Case 3

13.72 13.67 13.62 13.57 13.52 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour Fig. 8. Variation in EV bus (5th bus) voltage for different operation strategies.

day. Also, through electrification of transportation inside the micro grid, system’s overall emissions are reduced by 218.4 kg CO2 considering average fuel efficiency of 3 km/L with conventional fossil fuelled (diesel) vehicle with emission factor of 2.73 kg CO2/L. While, replacing conventional vehicle with EV saves diesel cost at an expense of electricity charge, the inclusion of EV’s into micro grid caused the shift in charging power of URFC. The URFC charging is shifted forward by 4 h (Fig. 4(b)) compared to case I where only URFC is scheduled to make the most of cheap electricity price. However, URFC is charged to its maximum capacity before discharged at peak electricity charges and also the rate and amount of energy discharged by URFC remained same as case I owing to high revenue attained at costly power tariff (Fig. 4(a) and (c)). The EVs are charged individually owing to their availability at grid, energy requirement, and power limits of the line. The EV1 is charged at 1st and 7th hours i.e., prior to the beginning of its transits (scheduled at hours 3 and 11). However, charging power of EV1 is more for 1st transit compared to 2nd and not at all charged

50

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

EV

PSO

URFC

Flowchart for micro grid operation strategies

Start

Initialization

Initialize all values, Itr=0,t=0, i=0 Generate Np particles with I dimension

Verification

Itr = Itr + 1

Verify all constraints equation

t=t+1 i=i+1

t Ѯ Ttrv No

Calculate Energy of URFC using equation

yes

U(I,t)=1

No

U(I,t)=0 No

UpdatePartical

Calculate fitness function

Load Flow

Energy Calculation

Calculate Energy of EV using equation

NO t==24

Yes

i==N

Yes

Forward Backward Sweep for Load Flow

Calculate objective function

Calculate Pbest and Gbest

Update particle using equation

Stopping criteria

Itr==Maxitr

Yes Stop

Flow chart 1. Operational strategies of micro grid with URFC and EV.

for third transit (Fig. 4(c)). This is due to the fact that, during transit all the energy is not consumed but the energy left over is used for next transit since V2G mode not enabled in this case. Similarly, EV2 also charged at 2nd, 4th and 6th hours i.e., prior to its transits

(scheduled at hours 3rd, 5th and 21st hours). It is also observed that, EV charging is scheduled at hours with cheap cost of electricity and in accordance with the network power constraints. The energy left in EV battery for next hours (time advanced by one

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

hour) is also presented (Fig. 4(g) and (h)) and energy dips represent beginning of EV transits. In the case also the power flow through PCC is within permissible limits. 5.3. Case 3: Micro grid operation with URFC and V2G operation In this case, battery storage of EV is harnessed to reduce the operational cost further through V2G operation in which EV’s are discharged at the time of peak/costly differential tariff. Intelligent scheduling of EV for V2G operation along with URFC reduced the objective function value i.e., cost of line loss to INR 8632.61. The V2G operation of EVs has resulted in savings of INR 1863 and 5228 compared to case I and case II respectively. However, savings from fuel conservation propels the net savings to INR 5991.5 per day compared to case I with conventional vehicle for transit inside micro grid. The URFC charging status is influenced by cost of electricity C(t) and load profile of micro grid. The charging power of EVs in case III (Fig. 5(c) and (d)) prior to first transit is considerably higher compared to case II (Fig. 4(c) and (d)) which can be attributed to the fact that, EV batteries are depleted to higher DOD due to V2G operation after 2nd transit and also EVs are not optimized to charge after V2G/prior to third transit. Also, during the V2G operation energy exchanged reduced after two hours so that EVs have sufficient residual energy for 3rd transit. The residual energy of EVs is shown in (Fig. 5(f) and (g)). It can be observed that, energy dip in EVs energy is associated with transit beginnings. The residual energy left after 24th hours the day is not zero unlike case II. Therefore the value of stored energy at the end of the day (Eq. (7)) is improved which will ultimately result in improved value of objective function Obj:case III (Eq. (25)) i.e., reduction in overall operation cost. However, there is no change in residual energy of URFC after 24th hour from case II to case III. The line losses of micro grid system are observed for every hour and it can be observed (Fig. 6) that, overall system losses depends on both load, distributed generation and charge and discharge profiles of storage elements. The peak active and reactive loads of the system are scheduled at hours between 8 and 13. However, there is a constant reduction in line loss from 8th to 13th hour due to increased amount of distributed generation of PV plants. Compared to case I, case II and case III have higher line losses at hours with low energy price (up to 8th hour) due to charging loads of EVs and URFC. The voltage profile of micro grid is also observed as in micro grid operation, inclusion of storage elements like URFC and EVs may affect system operating conditions causing violation of power quality standards. The maximum variation of 1.689%, 1.712% and 1.736% in case I, II and III respectively (Fig. 7) is observed which is well within allowable tolerance limits of ±5% deviation from rated voltage [31]. Also, the voltage profile of 5th bus is observed (Fig. 8). The maximum variations in 5th bus voltage for case I, II and III are 1.84%, 1.86% and 1.89% respectively which are within the tolerance limits of power quality standards. However, voltage variation of EV bus is greater than that of average system voltage. 6. Conclusion In this paper, an 11 bus micro grid network is simulated with distributed resources and storage elements URFC, EVs. The micro grid is operated under several network constraints like load balance, line capacity constraints and element limits/constraints like voltage, energy, power limits. For the given load and generation available, URFC and EV are scheduled intelligently to minimize overall multi objective function with objectives like line loss minimization and maximization of residual energy of storage elements

51

at end of the day. To optimize the schedule of storage elements, PSO with forward/backward sweep algorithms is used to solve power flow and cost minimization problems simultaneously through charge, discharge scheduling of URFC and EV. The EV is scheduled for two cases: charging only case and V2G case. The overall objective function value is least in case of V2G operation and resulted in a saving of INR 5228 and 5991.5 compared to micro grid operation strategies without V2G and conventional electric vehicle respectively. Combined with PSO, forward/backward sweep algorithm, EV scheduling in micro grids enabling V2G operation along with URFC can be a promising option for cost reduction in micro grids. Further inclusion of uncertainty prediction models, state estimation models respectively for EV mobility, SOC estimation can improve the work scheduling decisions, which may be considered in the future works.

References [1] Farhangi H. The path of the smart grid. IEEE Power Energ Mag 2010;8 (1):18–28. January–February. [2] Multi-year program plan 2011–2015 vehicle technologies program, energy efficiency and renewable energy, U.S. Department of Energy.

52

L.K. Panwar et al. / Energy Conversion and Management 106 (2015) 41–52

[22] Eberhart RC, Kennedy J. Particles swarm optimization. In: Proceedings of IEEE international conference on neural network, Piscataway, NJ, 1995. p. 1942–8. [23] Trovão João P, Antunes Carlos Henggeler. A comparative analysis of metaheuristic methods for power management of a dual energy storage system for electric vehicles. Energy Convers Manage 2015;95:281–96 [1 May]. [24] Eminoglu U, Hocaoglu MH. Distribution systems forward/backward sweepbased power flow algorithms: a review and comparison study. Electric Power Compo Syst 2009;37(1):91–110. [25] Bhansali Gaurav, Singh Bhanu Pratap, Kumar Rajesh. Strategic optimisation of microgrid by evolving a unitised regenerative fuel cell system operational criterion. Int J Sustain Energy 2014. http://dx.doi.org/10.1080/14786451. 2014.941367. [26] Xiaosong Hu, Li Shengbo, Peng Huei. A comparative study of equivalent circuit models for Li-ion batteries. J Power Sources 2012;198:359–67. January.

[27] Xiaosong Hu, Li Shengbo, Peng Huei, Sun Fengchun. Robustness analysis of State-of-Charge estimation methods for two types of Li-ion batteries. J Power Sources 2012;217:209–19. November. [28] Shirmohammadi D, Hong HW, Semlyen A, Luo GX. A compensation based power flow method for weakly meshed distribution and transmission networks. IEEE Trans Power Syst 1988;3(2):753–62. [29] BYD Electric Bus.