Generation scheduling in smart grid environment using global best artificial bee colony algorithm

Electrical Power and Energy Systems 64 (2015) 260–274

Contents lists available at ScienceDirect

Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes

Generation scheduling in smart grid environment using global best artificial bee colony algorithm Manisha Govardhan, Ranjit Roy ⇑ Department of Electrical Engineering, S.V. National Institute of Technology, Surat 395007, Gujarat, India

a r t i c l e

i n f o

Article history: Received 16 September 2013 Received in revised form 3 July 2014 Accepted 6 July 2014

Keywords: Distributed energy resources (DERs) Demand response program (DRP) Global best artificial bee colony (GABC) algorithm Gridable vehicle (GV) Unit commitment problem (UCP) Weibull distribution function

a b s t r a c t Generation scheduling is an important concern of the current power system which is suffering from many obstacles of limited generation resources, grown energy demand and fuel price, inconsistent load demand and fluctuations of available wind power in case of the thermal–wind system. Smart grid system has a great potential of tumbling existing power system difficulties with intelligent infrastructure and computation technologies. Three different distributed energy resources, namely, distributed generation, demand response and gridable vehicles are used in this paper to overcome the power system hitches. The classical generation scheduling is solved with insertion of the cost of demand response and the cost model pertaining to underestimation and overestimation of fluctuating wind power. The modified optimization problem is solved using an efficient Global best artificial bee colony algorithm for 10 generating units test system. Generation scheduling in the smart grid environment yields a significant reduction in the total cost. Ó 2014 Elsevier Ltd. All rights reserved.

Introduction Future power system has to be competent to deal with growing energy demand, global environmental alarm and inconsistent load demand, which are the major concern of the current power system. Extensive penetration of distributed energy resources (DERs) and their appropriate handling in the scheduling of generating units could be the probable solution to lessen the existing power system difficulties. Three types of DERs such as distributed generation (DG), customer participation in demand response program (DRP) as a responsive load and vehicle to grid (V2G) as an energy storage unit are considered in this paper [1,2]. V2Gs can feed power to the grid by discharging the battery and also referred to as gridable vehicles (GVs) [3]. Traditionally generation scheduling (usually referred to as unit commitment) is a mixed integer, combinatorial two linked optimization problem which decides when to start-up and shut down the generating units (unit commitment) and how to dispatch committed generators over a scheduled time horizon (economic load dispatch) to minimize the operating cost while satisfying the load demand and multiple constraints. Many researchers have developed several optimization techniques to solve unit commitment problem (UCP). The traditional methods include priority list method (PL) [4,5], branch and bound

⇑ Corresponding author. Tel.: +91 9904402937; fax: +91 261 2227334. E-mail address: [email protected] (R. Roy). http://dx.doi.org/10.1016/j.ijepes.2014.07.016 0142-0615/Ó 2014 Elsevier Ltd. All rights reserved.

method (BB) [6], dynamic programming (DP) [7], mixed-integer programming (MIP) [8] and Lagrangian Relaxation (LR) [9,10]. The classical PL method is simple and fast but yields higher generation cost. DP and BB suffer from the dimensionality problem which results in excessive computation time as the number of generating unit increases. LR method has convergence problem and generates poor quality solution. Recently, some methods based on meta-heuristics approach are also available such as genetic algorithm (GA) [11,12], evolutionary programming (EP) [13], simulated annealing (SA) [14], fuzzy logic (FL) and particle swarm optimization (PSO) [15–18], tabu search [19] and ant colony optimization (ACO) [20]. These methods can execute complex problems with high quality resolution and can reach up to or near the global optimal solution. However, with the large-scale unit system all these algorithms adversely result in enormous computation time and some methods yield suboptimal solutions. With this perspective, recently established global best artificial bee colony algorithm (GABC) [21,22] is implemented to solve UCP to achieve the global optimum solution. Moreover, several researches have been carried out to solve UCP with demand response and GVs individually to achieve an economic and ecological solution. The model of emergency demand response program (EDRP) and interruptible load contracts (ILC) in UCP is proposed in [23] to minimize the energy consumption during the critical or peak period of the day. Another UCP associated with demand response is suggested in [24] to study the environment and economic effect.

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

261

GVs have great potential of emission reduction and reliability enhancement. The intelligent, cost-effective and eco-friendly study of UCP with GVs is discussed in [25]. Also, GVs can reshape the load demand by charging from the grid during off-peak hours and feeding (discharging) power to the grid during peak hours. Two novel techniques of load levelling and smart grid models of GVs for cost and emission reduction with UCP are presented in [26]. The vehicle to grid algorithm has been developed in [27] to optimize the scheduling of energy and ancillary services. The analysis of electric vehicle mobility in transmission constrained generation scheduling is given in [28]. Some research literature on the collective impact of demand response and GVs on generation scheduling are also published recently. In [2], reliability constrained UCP is solved considering the cost of demand response and GVs in the objective function. Moreover day-ahead resource scheduling with demand response and GVs is discussed in [29,30]. This paper has modified classical UCP formulation with the integration of the cost model of uncertain wind power in terms of overestimation and underestimation cost along with the cost of DRP. However, the cost model for economic load dispatch which considers overestimation and underestimation cost of wind generation has been discussed in [22,31], but in this paper the same concept is stretched for UCP. According to [32], among the different demand response models (linear, potential, exponential and hyperbolic), hyperbolic and potential demand models are unable to attain the optimum hourly solution. The most reasonable linear function model for UCP has been developed in [24]. Hence, this paper has chosen another moderate approach to exponential load economic model for exhaustive study of demand response. A well-known emergency demand response program (EDRP) is considered in this paper. The main contribution of this work is as follows:

Federal Energy Regulatory Commission (FERC) order 719 has classified different DRPs into two main categories, namely, incentive based programs and time based programs [33]. The incentive based DRP offers cash or discount in bill to the customers for reducing their electricity consumption during peak hours or during periods of high electricity price. The well-known time based DRP motivates the customers to shift their load from peak hours to low load or off-peak hours. DRP has been proved as an organized approach to reshape the inconsistent load demand curve by reducing the load demand during peak period of the day in [34,35]. Electric vehicles with the competence of feeding power to the grid are usually referred to as gridable vehicles (GVs) and can be served as DER when GVs are parked. GVs charged during off-peak hours can insert power to the grid during peak hours which in turn results in peak hour saving. Occurrence of GV reduces the CO2 emission and enhances the system reliability. This paper has assumed three individual DER aggregators to communicate with the independent system operator (ISO). ISO will dispatch power according to the information gathered from these aggregators as shown in Fig. 1. Wind aggregator will collect information of wind power generation from wind farm owner and accordingly decide the associated overestimation and underestimation cost. DR aggregator will offer various incentive schemes to the customer for peak load reduction and keeps a record of the participating customer through smart metering. GV aggregator will maintain a history of registered vehicles, their grid connected timings and depth of charge level. This study has assumed the smart grid system which can assure appropriate operation and control of these DERs with the intelligent infrastructure and computational technologies.

 The conventional UCP objective function is modified by including the direct cost, overestimation and underestimation cost of uncertain wind power generation.  The exponential demand function is solved to comprehend the impact of EDRP on generation scheduling.  The entire optimization problem is solved considering the impact of EDRP and GVs on generation scheduling along with uncertain wind power using an efficient global best artificial bee colony (GABC) algorithm.

Wind probability distribution function Uncertainty of wind speed can be modelled from the probability distribution function of wind power. It has been already proved that the Weibull distribution function [31,36] is proficient to characterize the uncertain wind speed and is given as:

The rest of the paper is deployed as follows: Section ‘Distributed energy resources’ deals with the essential framework of DERs considered in this paper. Section ‘UC problem formulation’ describes the modified UCP along with the cost model of DRP and uncertain wind power. A brief structure of the GABC optimization algorithm and its implementation is described in Section ‘Overview of global best artificial bee colony optimization algorithm’. Section ‘Results and discussion’ deliberates the simulation results and comparison of different scenarios considered. The conclusion is drawn in Section ‘Conclusion’.

Mathematical modelling

f ðv Þ ¼

  ðk1Þ   k  k v v 0
ð1Þ

where function f(v) gives the time span in which wind has a velocity of v m/s, c and k denote scale factor and shape factor of a particular location respectively. The probability of having wind velocity equal to or less than v m/s is defined as the cumulative distribution function and is described as:

  k  v Fðv Þ ¼ 1  exp  c

ð2Þ

After mapping the inconsistent wind speed as a random variable, the wind power generated from the wind power generator [31] is calculated as:

v < v ci and v > v co ðv  v ci Þ PW ¼ PWr ; v ci 6 v 6 v r ðv r  v ci Þ PW ¼ PWr ; v r 6 v 6 v co PW ¼ 0;

Distributed energy resources Three types of DERs are considered in this paper, namely, distributed generation (DG), emergency demand response program (EDRP) and gridable vehicles (GVs). DG plays a significant role in the power system generation and operation. Upcoming wind energy has been identified as a new challenge in distributed generation due to unpredictable wind nature. This paper has considered a wind power generator as DG and its consequence on generation scheduling.

ð3Þ

where PW and PWr are output and rated power of wind power generator respectively whereas vci, vco and vr are signified as cutin, cut-out and rated speed of wind power generator respectively. From the Weibull probability distribution function [31], the probability of wind power being zero, rated and between zero and rated power can be formulated as (4)–(6) respectively:

262

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Fig. 1. Framework of generation scheduling with DERs.

  k    k  v ci v co P r fP W ¼ 0g ¼ Fðv ci Þ þ ð1  Fðv co ÞÞ ¼ 1  exp  þ exp  c c

ð4Þ   k    k  vr v co  exp  Pr fPW ¼ PWr g ¼ Fðv co Þ  Fðv r Þ ¼ exp  c c ð5Þ Pr f0 < P W < PWr g ¼

 k1  k ! klv ci ð1 þ qlÞv ci ð1 þ qlÞv ci exp  c c c ð6Þ

where q = PW/PWr and l = (vr  vci)/vci. Demand response Demand response is a customer driven incentive based program in which the participating customers are rewarded by cash or discount in their electricity bill for reducing their electricity usage during peak hours or shifting their flexible load to low load and off-peak hours. In deregulated electricity market the customer participation can be assessed from the economic model of load demand which reveals the change in load demand with the change in price of electricity and the amount of incentive offered to them during several periods of the day. This demand elasticity (E) pertaining to electricity price (EP) [35] is defined as:

   EP o @PL E¼ PLo @EP

ð7Þ

where EPo and PLo are initial electricity price and load demand respectively. oEP and oPL describe the change in electricity price and load demand from their initial values respectively. Customer behavior is characterized according to the load variation with the change in the electricity price. There are certain inflexible loads which cannot shift from one period to another with the price variation and are sensitive to single period only. These loads are termed as self-elasticity. Furthermore, some elastic loads that can vary from

peak hours to low load periods having sensitivity to multi-period can be defined as cross elasticity. Consequently, customer behavior for 24 h can be epitomized by price elasticity matrix (PEM) which is a 24  24 matrix with self-elasticity coefficients as diagonal elements and cross elasticity coefficients as off-diagonal elements [35]. In DRP, the participating customers change their load demand according to the incentive value (A) offered to them. The incentive paid to the customer in hth period for reduction of each kW h load demand is given as [35]:

IðDPLðhÞ Þ ¼ AðhÞ ½PLoðhÞ  PLðhÞ 

ð8Þ

where PL signifies the load demand after implementing DRP. Total profit (TB) of the customer in hth hour with DRP can be written as [35]:

T B ¼ BðPLðhÞ Þ  PLðhÞ EPðhÞ þ IðDPLðhÞ Þ

ð9Þ

where B(PL(h)) is the customer’s income during hour h and EP(h) is the electricity price at hth hour. The maximum benefit can be attained by making oTB/oPL(h) = 0 which results in:

@BðPLðhÞ Þ ¼ EP ðhÞ þ AðhÞ @ðPLðhÞ Þ

ð10Þ

Various profit functions are discussed in [32] for demand side management. This paper has considered the exponential demand function which can be characterized as:

     PLðhÞ 1 1 BðP LðhÞ Þ ¼ BoðhÞ þ EP oðhÞ PLðhÞ 1 þ ln EðhÞ PLoðhÞ

ð11Þ

where Bo(h) is initial income of the customer. Differentiation of (11) with respect to PL(h) results in:

   @BðPLðhÞ Þ PLðhÞ 1 ¼ EP oðhÞ 1 þ ln EðhÞ PLoðhÞ @PLðhÞ

ð12Þ

Substituting (12) into (10) yields

   PLðhÞ 1 EPðhÞ þ AðhÞ ¼ EP oðhÞ 1 þ ln EðhÞ PLoðhÞ

ð13Þ

263

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Redeploying (13) provides the participating customer’s consumption as:

  EP ðhÞ  EP oðhÞ þ AðhÞ PLðhÞ ¼ PLoðhÞ exp EðhÞ EPoðhÞ

ð14Þ

UC problem formulation This section describes the mathematical model of modified UCP. The objective of the proposed model is to minimize the total cost (TC) while satisfying typical constraints of traditional unit commitment over the scheduled horizon.

TC ¼ MIN

NG X T X W f ðFC i ðPði;hÞ Þ þ SC ði;hÞ ð1  uiðh1Þ ÞÞ þ W e Emi ðP ði;hÞ Þ i¼1 h¼1

ð15Þ where Wf and We are weighing factors which decide insertion and omission of the fuel cost and emission into the objective function respectively. NG is specified as the number of generating units. T is the schedule time horizon and ui indicates the on/off status of ith unit and h is the hour index. The operation cost of a system is an accumulation of fuel cost and start-up cost of all committed units over the scheduled time span. Fuel cost (FC) of a generating unit in quadratic polynomial form is defined as:

FC i ðPði;hÞ Þ ¼ ai þ bi P ði;hÞ þ ci P2ði;hÞ

ð16Þ

where ai, bi and ci are the fuel cost coefficients and P(i,h) is the power output of ith generating unit at hour h. Usually, the start-up cost depends on boiler temperature of a thermal unit and is specified as:

SC ði;hÞ ¼



U off ði;hÞ

6 MDi þ CSHi HSC if MDi 6 off CSC if U ði;hÞ > MDi þ CSHi

ð17Þ

ð18Þ

where ai, bi and ci are the emission coefficients of unit i. This paper has incorporated the cost of uncertain wind power and the cost of demand response program along with the traditional UC problem. Inspired by [37], overall objective function (OBF) with one of the functions (FC or Em) to be minimized can be rewritten as:

OBF ¼ TC þ

NW X C Wði;hÞ þ C DRðhÞ

C dw;i ðPWði;hÞ Þ ¼ di PWði;hÞ

ð21Þ

where d is the direct cost of each wind power generator in $/h.  Underestimation cost function If the available wind power is higher than the expected wind power, the ISO has to pay a penalty to the wind farm owner for not using all available wind power. The penalty or underestimation cost [31] for waste of wind power is considered as:

C pw;i ðPW;av lði;hÞ  PWði;hÞ Þ ¼ kp;i ðP W;av lði;hÞ  PWði;hÞ Þ Z PWr;i ¼ kp;i ðPW  PWi Þf ðPW ÞdPW

ð22Þ

PW;i

where kp is the penalty cost coefficient in $/h and f(PW) is the wind power probability distribution function.  Overestimation cost function If wind power generated is less than the expected wind power, ISO has to purchase extra wind energy from another source for reliable system operation by paying higher price. Hence, overestimation cost [31] is simply a reserve cost paid by the ISO and it can be formulated as:

C rw;i ðPWði;hÞ  PW;av lði;hÞ Þ ¼ kr;i ðPWði;hÞ  PW;av lði;hÞ Þ Z PW;i ¼ kr;i ðP Wi  P W Þf ðPW ÞdPW

ð23Þ

0

where U(i,h)off is the period during which ith unit remains continuously off up to hour h. HSC and CSC are referred to as hot and cold start-up cost respectively. MDi is minimum down time and CSHi is cold start hour of unit i. Like fuel cost, emission pollution (Emi) of ith generating unit can also be expressed in polynomial form as:

Emi ðPði;hÞ Þ ¼ ai þ bi Pði;hÞ þ ci P2ði;hÞ

 Direct cost function Generally a wind turbine or a wind farm (cluster of wind turbines) is kept by some private entity and ISO would have to pay a certain amount under predetermined contracts for wind power purchase from a private body. The direct cost of generating wind power [31] is assumed to be a linear function which is described as:

ð19Þ

where kr is the reserve cost coefficient in $/h. Usually, values of direct cost (d), reserve cost coefficient (kr) and penalty cost coefficient (kp) are decided independently by the ISO which may be different for different regions. Moreover, the cost of demand response is an amount paid by the ISO to the customers for their participation in DRP which is largely influenced by the incentive value and PEM. In EDRP, the incentive is offered during peak hours only and during low load and off-peak hours it is considered as zero to improve the load profile during peak hours. Again, the incentive value offered to the customers is completely ISO’s decision depending on peak levelling, customer participation rate and PEM. Hence, the cost of demand response [23] is the amount of incentive paid to the customers for their active participation in DRP and can be calculated as:

C DRðhÞ ¼ IðDPLðhÞ Þ ¼ AðhÞ ðP LoðhÞ  PLðhÞ Þ

ð24Þ

The objective function (19) must satisfy the following constraints:

i¼1

where NW specifies the number of wind power generator. The cost model of wind power generator [31] implicates the direct cost along with the underestimation and overestimation cost of unpredictable wind power as:

C Wði;hÞ

NW NW X X ¼ C dw;i ðPWði;hÞ Þ þ C pw;i ðPW;av lði;hÞ  PWði;hÞ Þ i¼1

þ

NG NW X X Pði;hÞ uði;hÞ þ PWði;hÞ þ PVðhÞ ¼ ð1  bÞPLoðhÞ i¼1

i¼1

NW X C rw;i ðPWði;hÞ  PW;av lði;hÞ Þ

A. Power balance constraints: Accumulation of power generated from thermal units, wind power generator and GVs must satisfy the load demand and is characterized as:

ð20Þ

i¼1

where Cdw, Cpw and Crw are the cost models of direct cost, underestimation cost and overestimation cost respectively. PW,avl is the wind power available from the wind power generator.

i¼1

  EPðhÞ  EP oðhÞ þ AðhÞ þ bPLoðhÞ exp EðhÞ EPoðhÞ

ð25Þ

where PV(h) is the power output of the vehicle at hour h and b is the customer participation rate (in percentage). If GVs are assumed to inject power to the grid all the time then PV(h) can be calculated as [25]:

264

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

PVðhÞ ¼ gPv ð1  wÞNVðhÞ

ð26Þ

where NV(h) is the number of vehicles connected to the grid at hour h whereas g, w and Pv are inverter efficiency, state of charge of battery and average capacity of GV respectively. B. Generation limit: Power should be generated within specified generation limits as: max Pmin ðiÞ 6 P ðiÞ 6 P ðiÞ

ð27Þ

where Pmin and Pmax are minimum and maximum generation limits of each thermal unit respectively. C. Minimum up/down time limit: Each unit has to remain on or off for a predefined time period before any change over takes place as:

U on i P MU i

ð28Þ

U off P MDi i

where Uon and Uoff are the periods during which generating units remain continuously on and off respectively whereas MU and MD are minimum up and down time of each generating unit respectively. D. Spinning reserve constraint: Adequate spinning reserve should be required for stable and reliable operation. Classically SR(h) is predefined as maximum generation capacity of a unit available in the system or specific percentage of load demand. Inspired by [3,25], the SR constraint can be designed as: NG NW X X Pmax PWði;hÞ þ P VðhÞ ¼ P LðhÞ þ SRðhÞ ði;hÞ uði;hÞ þ i¼1

ð29Þ

i¼1

E. Ramp rate: Difference between power generations from one hour to another should be within the specified ramp up and ramp down limits as:

Pði;hþ1Þ  Pði;hÞ 6 RU i Pði;hÞ  P ði;hþ1Þ 6 RDi

ð30Þ

where RU and RD are ramp up and ramp down limits of each unit respectively. F. Initial status (IS): Initial status of each unit should be considered before commencement of the scheduling. G. Vehicle balance: Only registered vehicles can feed power to the grid and summation of all vehicles should be equal to the total number of registered vehicles (N total ) after schedV uled horizon. T X NVðhÞ ¼ Ntotal V

ð31Þ

h¼1

H. State of charge (SOC) (w): Each vehicle should hold expected SOC at all the time. I. Discharging frequency: GVs are assumed to be charged from renewables and fed power to the grid. Multiple discharging frequencies of GVs may be possible but for the sake of simplicity, each vehicle is assumed to discharge once a day [25]. J. Vehicle limit: For system reliability, the number of vehicles connected to the grid per hour is fixed.

NVðhÞ 6 Ntotal V

ð32Þ

K. Efficiency (g): For practical approach, battery efficiency should be considered. Overview of global best artificial bee colony optimization algorithm In 2005, Karaboga has developed the elementary concept of artificial bee colony algorithm (ABC) which emphases on intelligence of honeybees in their food finding habits and social

interactions with each other. The artificial bees are mainly distributed into three groups, namely, employed bees, onlookers and scouts [38]. These bees fly in multidimensional search space to pursue their food source. The bees that use their own experience to find food source are called employed bees. Onlooker bees are waiting on hive to gather food source information during the waggle dance from employed bees. Some bees hunt their food sources arbitrarily are called scout bees. The onlooker bees pick good food source from the information conveyed by employed bees by evaluating the probability [38] of a food source using (33) as:

, pi ¼ fit i

Ne X fit j

ð33Þ

j¼1

where fit is the fitness value of ith solution and is proportional to the nectar amount of food source. Ne is the number of employed bees which is equal to the number of food source. Then, the onlooker bees further search new food source vij [38] nearby the selected food source using (34) as:

v ij ¼ Pij þ /ij ðPij  Pkj Þ

ð34Þ

where Pij is the previous food source invented by employed bees, Pkj indicates the alternative food source chosen in neighborhoods by onlooker bees, /ij denotes a random number between 1 to 1, j € {1,2,. . .D} and k € {1,2. . .Ne} are random number index subject to k – i. D is dimension of the search space and /ij controls the formation of a food source position in neighborhood. After comparing nectar amounts of previous (Pij) and new food source (vij), onlooker bees remember the position of the best food source among the two into their memory. If the solution pertained to a food source is not upgraded up to certain trials, then that food source is rejected and associated bee converted into scout bee. The number of trials for eliminating the food source is referred to as a limit value which is an important control parameter of ABC algorithm. Then the scout bee searches for new food source randomly using (35) as:

Pij ¼ Pmin þ randðPmax  Pmin Þ j j j

ð35Þ

where rand is the random number between 0 and 1. Pmin and Pmax j j describe the minimum and maximum limits of parameter to be optimized respectively.  Global best artificial bee colony algorithm (GABC) In an optimization algorithm, exploration and exploitation must be carried out together in the dynamic search process. Exploration is referred to as an ability to find the global optimum solution in entire search space and exploitation means an ability to employ previous experience of good solution to attain better solution. Since, basic ABC is good at exploration and bad at exploitation, Eq. (35) is modified to enhance exploitation capability as [21]:

v ij ¼ Pij þ /ij ðPij  Pkj Þ þ Wij ðyj  Pij Þ

ð36Þ

The added term in (36) is the Gbest term, yj is the jth element of the global best solution and Wij is a random number in [0, C] where C is a nonnegative constant. Implementation of GABC in proposed UCP The generalized algorithm steps for solving proposed UCP using GABC are as follows: Step 1: Define system data like number of generators, maximum and minimum generation capacities, cost coefficients, minimum up and down time and initial status of each generator, scheduled period and corresponding load demand, price elasticity matrix and all necessary data for GVs and wind power generator system.

265

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Fig. 2. Flowchart of GABC to solve proposed UC problem.

Table 1 10-Unit system data. Unit

Pmax (MW)

Pmin (MW)

a ($)

b ($/MW h)

c ($/MW h2)

MU (h)

MD (h)

RU (MW)

RD (MW)

HSC ($)

CSC ($)

CSH (h)

IS (h)

U1 U2 U3 U4 U5 U6 U7 U8 U9 U10

455 455 130 130 162 80 85 55 55 55

150 150 20 20 25 20 25 10 10 10

1000 970 700 680 450 370 480 660 665 670

16.19 17.26 16.60 16.50 19.70 22.26 27.74 25.92 27.27 27.79

0.00048 0.00031 0.00200 0.00211 0.00398 0.00712 0.00079 0.00413 0.00222 0.00173

8 8 5 5 6 3 3 1 1 1

8 8 5 5 6 3 3 1 1 1

152.5 152.5 55.0 55.0 68.5 30.0 30.0 22.5 22.5 22.5

152.5 152.5 55.0 55.0 68.5 30.0 30.0 22.5 22.5 22.5

4500 5000 550 560 900 170 260 30 30 30

9000 10,000 1100 1120 1800 340 520 60 60 60

5 5 4 4 4 2 0 0 0 0

8 8 5 5 6 3 3 1 1 1

Step 2: Initialization: Initial population P = [P1, P2, P3,. . ., Pn]T of arbitrarily distributed solution is generated from the multidimensional search space, where vector of P1, P2, P3,. . ., Pn is

the possible candidate solutions and n is the size of optimized parameters. Each solution vector is specified as P i ¼ ½P i1 ; Pi2 ; Pi3 ; . . . ; P iNG ; P Wi1 ; . . . ; PWiNW  where i e [1, n].

266

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Table 2 Hourly load demand and spot electricity price. Hour

1

2

3

4

5

6

7

8

9

10

11

12

Load demand (MW) EP ($) Hour Load demand (MW) EP ($)

700 22.15 13 1400 24.6

750 22 14 1300 24.5

850 23.1 15 1200 22.5

950 22.65 16 1050 22.3

1000 23.25 17 1000 22.25

1100 22.95 18 1100 22.05

1150 22.5 19 1200 22.2

1200 22.15 20 1400 22.65

1300 22.8 21 1300 23.1

1400 29.35 22 1100 22.95

1450 30.15 23 900 22.75

1500 31.65 24 800 22.55

Table 3 Price elasticity matrix.

Low Off-peak Peak Off-peak Peak

Low

Off-peak

Peak

Off-peak

Peak

0.08 0.03 0.034 0.03 0.034

0.03 0.11 0.04 0.03 0.04

0.034 0.04 0.19 0.04 0.01

0.03 0.03 0.04 0.11 0.04

0.034 0.04 0.01 0.04 0.19

Step 3: Implement load economic model of (14) and generate a new set of load demand with the assumed customer participation, incentive value and price elasticity matrix. Step 4: Randomly generate GV connected to the grid and calculate vehicle power using (26). Step 5: Each particle of population Pi is distributed uniformly between its maximum and minimum specified generation limits according to (35) where i and j denote particle number and generating unit respectively. Then, calculate the required power to be generated (i.e. PG = PL  PV) while satisfying the load balance and reserve constraints and solve the UCP by distributing PG among the committed units and wind power generator. Step 6: Fitness calculation: Initially for cycle = 1, calculate the fitness value of each food source pertaining to employed bees and repeat the procedure until the termination criteria (maximum number of cycles) has been reached. Step 7: Employed bee phase: In this phase, each employed bee generates a new food source in the neighborhood of its current position employing (36). All the constraints (25)–(32) are checked for the new food source position. Then fitness value of the new food source is calculated and compared with the previous one. If the fitness value of the new food source is better than the previous one, then it is remembered otherwise the previous solution is considered. Step 8: Probability calculation: Onlooker bees select better food source position found by the employed bee from the probability calculation using (33). Step 9: Onlooker phase: In this phase, all onlooker bees further hunt for better food source in the neighborhood of the selected food source according to (36). Again, food source position with better fitness value is remembered and worse is eliminated. Step 10: Scout phase: If the solution is not improved up to certain trial, then the onlooker bee converted into scout bee and create the new random food source using (36) to attain better solution. Step 11: The best food source position up till now is remembered and then increment the cycle number.

Fig. 3. Total cost for different colony size.

Fig. 4. Total cost for different onlooker bee. Number of onlooker bees = (0, 2, 4,. . .10) * N.

Step 12: Stop after reaching the stopping criteria (maximum number of cycles) otherwise go to step 7. The flow chart of GABC algorithm to solve proposed UCP is given in Fig. 2. Results and discussion This section involves an exhaustive study on diverse test cases of UCP performed on a typical 10 unit system. Table 1 [39] compiles all the essential details of generating units and Table 2 [40] carries hourly load demand with associated spot electricity price. This paper has considered equal value of electricity price before and after implementation of DRP. Desired price elasticity matrix shown in Table 3 for implementing DRP can be found in [24]. The emission coefficients of generating units are given in Table 4 [25]. The entire load demand curve is divided into various periods of the day, namely, low load period (00.00–05.00 h), off-peak

Table 4 Emission coefficients. Unit

a (ton/h)

b (ton/MW h)

c (ton/MW2 h)

Unit

a (ton/h)

b (ton/MW h)

c (ton/MW2 h)

U1 U2 U3 U4 U5

10.33908 10.33908 30.03910 30.03910 32.00006

0.24444 0.24444 0.40695 0.40695 0.38132

0.00312 0.00312 0.00509 0.00509 0.00344

U6 U7 U8 U9 U10

32.00006 33.00056 33.00056 35.00056 36.00012

0.38132 0.39023 0.39023 0.39524 0.39524

0.00344 0.00465 0.00465 0.00465 0.00470

267

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

x 105

9

Total cost ($)

8.5 8 7.5 7 6.5 6 5.5

Fig. 5. Total cost for different limit value.

0

50

100

150

200

Number of cycles Fig. 9. Convergence characteristics for Wf = 1, We = 0.

x 105 8

Fig. 6. Total cost for different value of constant C.

Emission (ton)

7 6 5 4 3 2

0

50

100

150

200

Number of cycles Fig. 10. Convergence characteristics for Wf = 0, We = 1.

x 106 3

Total cost ($)

2.5 Fig. 7. Total cost comparison for Wf = 1, We = 1 and Wf = We = 0.5.

2 1.5 1 0.5

0

50

100

150

200

Number of cycles Fig. 11. Convergence characteristics for Wf = 0.5, We = 0.5.

Fig. 8. Emission comparison for Wf = 1, We = 1 and Wf = We = 0.5.

given in [25]. The specifications are: total number of vehicles, NV total = 50,000; average battery capacity, Pv = 15 kW h, number of vehicles connected to the grid per hour = 10% of Ntotal , efficiency, V g = 85%, desired state of charge, W = 50%. Selection of simulation parameters of GABC

period (05.00–09.00 h and 14.00–19.00 h) and peak load period (09.00–14.00 h and 19.00–00.00 h) respectively. This study has considered 10% spinning reserve for system reliability and assumed 40% customer participation in DRP according to [24]. The industrial customers can utilize DRP at great extent due to the flexibility of shifting their heavy load from peak to off-peak period while residential customers with low load demand have less impact on DRP. As a closer view of a city or a state, commercial customers are more than industry and hence 40% participation is justified. This paper acknowledges same specifications of GVs

GABC algorithm has been implemented to solve UCP after appropriate selection of different control parameters such as colony size, number of onlooker bees, limit and nonnegative constant C for the optimum solution with less computational time. These parameters are selected by varying single parameter keeping other parameters constant after 20 simulation trials. To evaluate the impact of colony size on the total cost, maximum number of cycles, limit value and C are fixed to 200, 2 and 1.5 respectively. Then colony size (N) is varied from 50 to 550 with equal intervals shown in

268

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Fig. 3 which shows reduction in total cost as colony size increases and minimum value is obtained at 350 colony size. Onlooker bees also have an influence on the optimum result, hence onlooker bees are varied from 0.5 * N to 10 * N. The increase in colony size and number of onlooker bees result in increased computation time. The plot of total cost for different onlooker bees is given in Fig. 4 which confirms that the optimum result is obtained when onlooker bees are equal to the twice of the colony size. The plots of total cost with different values of limit and nonnegative constant C are shown in Figs. 5 and 6 respectively which yield minimum result at limit = 2 and C = 1.5. Hence, optimization parameters selected for GABC to solve proposed optimization problem are: colony

size = 350, number of onlooker bees = 2 * 350, limit = 2, C = 1.5 and maximum number of cycles = 200. Case studies This paper investigates different case studies of UCP without and with integration of uncertain wind power in smart grid environment. Case 1: To comprehend the effect of smart grid deployment on UCP in uncertain wind environment, it is important to accomplish economic-environmental study. Hence, UCP has been

Table 5 Generation scheduling for base case. Hour

U1 (MW)

U2 (MW)

U3 (MW)

U4 (MW)

1 455 244.8 0 0 2 455 295.0 0 0 3 455 361.6 0 0 4 455 452.2 0 0 5 455 455.0 54.5 0 6 455 455.0 109.5 49.4 7 455 423.4 130.0 103.1 8 455 455.0 95.1 128.4 9 455 455.0 116.2 130.0 10 455 455.0 107.8 130.0 11 455 455.0 129.8 130.0 12 455 454.7 124.6 130.0 13 455 450.1 126.5 128.4 14 455 455.0 71.5 130.0 15 455 425.9 112.7 129.0 16 455 366.0 122.7 81.3 17 455 365.2 68.6 86.1 18 455 373.1 104.1 129.5 19 455 430.7 130.0 124.0 20 455 455.0 130.0 130.0 21 455 447.4 129.5 130.0 22 455 450.4 0 0 23 455 362.3 0 0 24 455 345.0 0 0 Total cost = $565,614, emission = 269091.3 ton

U5 (MW)

U6 (MW)

U7 (MW)

U8 (MW)

U9 (MW)

U10 (MW)

FC ($)

SC ($)

Emission (ton)

0 0 33.4 42.8 35.5 31.1 38.5 66.5 88.4 146.0 160.0 158.0 139.0 144.0 77.4 25.0 25.0 38.3 60.3 129.0 60.3 127.0 82.7 0

0 0 0 0 0 0 0 0 30.0 60.0 64.3 77.4 47.4 20.0 0 0 0 0 0 30.0 52.8 22.8 0 0

0 0 0 0 0 0 0 0 25.4 25.0 25.0 40.1 32.5 25.0 0 0 0 0 0 30.0 25.0 45.1 0 0

0 0 0 0 0 0 0 0 0 21.7 10.0 32.5 21.1 0 0 0 0 0 0 22.5 0 0 0 0

0 0 0 0 0 0 0 0 0 0 20.7 10 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 17.5 0 0 0 0 0 0 0 0 0 0 0 0

13679.59 14554.50 16829.92 18604.74 20118.33 22461.87 23309.08 24261.08 27321.80 30256.94 31965.60 34005.10 30230.15 27462.24 24281.53 21541.85 20691.67 22430.63 24229.89 29761.90 27355.73 22896.05 17836.38 15427.42

0 0 900 0 550 1120 0 0 860 60 60 60 0 0 0 0 0 0 0 490 0 0 0 0

6827.6 7547.8 8989.1 11042.8 11357.1 11820.3 11378.8 12105.5 12794.5 13226.7 13815.9 14045.6 13173.6 12803.6 11517.8 9959.3 9653.1 10266.9 11721.0 13320.6 12640.3 11661.4 8997.95 8424.02

Table 6 Generation scheduling with GVs. Hour

U1 (MW)

U2 (MW)

U3 (MW)

U4 (MW)

U5 (MW)

1 455 237.18 0 0 0 2 455 287.39 0 0 0 3 455 335.96 0 0 54.50 4 455 426.29 0 0 61.65 5 455 404.91 0 0 108.60 6 455 445.24 51.52 0 135.30 7 455 452.44 105.07 55.00 71.00 8 455 451.55 129.02 96.21 62.83 9 455 455.00 129.41 129.55 96.37 10 455 455.00 126.43 130.00 161.90 11 455 453.49 128.75 130.00 158.50 12 455 455.00 130.00 129.91 162.00 13 455 455.00 126.74 127.32 111.50 14 455 454.78 118.87 130.00 90.54 15 455 455.00 93.07 129.01 49.85 16 455 302.50 82.55 125.16 69.51 17 455 337.66 107.40 70.16 25.00 18 455 395.43 94.38 97.99 32.65 19 455 448.72 129.95 126.09 32.71 20 455 441.10 125.58 111.81 120.80 21 455 444.20 105.91 129.06 115.80 22 455 432.67 0 130.00 47.28 23 455 417.11 0 0 0 24 455 338.55 0 0 0 Total cost = $552110.98, emission = 263754.6 ton

U6 (MW)

U7 (MW)

U8 (MW)

U9 (MW)

U10 (MW)

No. of vehicles

Emission (ton)

GV demand (MW)

GV capacity (MW)

0 0 0 0 0 0 0 0 0 30.00 47.72 77.72 80.00 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 30.00 25.00 25.00 25.00 29.78 42.86 0 0 0 0 0 68.10 44.04 25.00 0 0

0 0 0 0 0 0 0 0 0 0 22.50 38.87 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 21.33 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1301 1193 711 1108 4937 2025 1802 845 733 2609 1206 4157 2301 1248 2834 2395 750 3852 1181 4817 944 1576 4375 1012

6726.91 7422.13 8469.13 10390.31 9994.42 11302.95 11710.66 12040.74 12680.51 13231.16 13716.45 13510.84 12920.27 12575.27 12100.69 8799.47 9250.78 10445.86 12205.09 12264.32 12271.36 11454.84 9959.78 8310.64

8.29 7.61 4.53 7.06 31.50 12.90 11.50 5.39 4.67 16.60 7.69 26.50 14.70 7.96 18.10 15.30 4.78 24.60 7.53 30.70 6.02 10.00 27.90 6.45

16.59 15.21 9.06 14.13 62.95 25.82 22.98 10.77 9.34 33.26 15.38 53.00 29.34 15.91 36.13 30.54 9.56 49.11 15.06 61.42 12.04 20.09 55.78 12.90

The italicized units remain off due to the presence of the GVs compared to the base case.

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

solved as the base case with the typical 10 unit system using GABC algorithm in this section. Initially, this case is solved to attain the economic solution without concerning about environmental pollution (Wf = 1, We = 0). Secondly, only the emission concern has been considered (Wf = 0, We = 1) and finally, both economic and emission objective functions are considered (Wf = 0.5, We = 0.5). Case 2: This case consolidates the influence of GVs on total cost and generation scheduling of thermal units. GVs are assumed to be charged from renewables and fed to the grid all the times. Case 3: This case scrutinises the impact of EDRP on generation scheduling and total cost with the change in customer participation rate, PEM and amount of incentives offered to the customers. EDRP offers high incentive value during peak load hours to encourage the customers to reduce their load demand to get financial benefit. The incentive offered to the customers for their participation in EDRP is independently decided by the ISO. This study assumes $10 as the peak hour incentive. Case 4: This case incorporates the cost of uncertain wind power in terms of the direct cost, underestimation and overestimation cost in the objective function of the traditional UCP. Case 5: This case deals with the combined effect of GVs and EDRP on the total cost and generation scheduling. Case 6: This case deliberates the collective effect of EDRP and GVs on UCP with indeterminate wind power integration.

269

and without GVs are shown in Figs. 12 and 13 respectively, which specify that GV improves the system reliability and reduces emission pollution. Case 3: DRP also has a great proficiency of tumbling emission pollution with the enriched load demand profile. Total cost including fuel, start-up and demand response cost with EDRP is $553479.2 which is $12134.8 (565614–553479.2) less than the base case. The simulation study is also carried out for double and half of the PEM elements. The load demand plots for double, base and half PEM elements are shown in Fig. 14. Scheduling of generated power in each unit for these three sub cases is shown in Fig. 15. It is clear from Fig. 15 that generating unit U10 remains on for half PEM elements and U9 remains on for both half and base PEM elements. While both the units U9 and U10 remain in off condition for double PEM elements which ultimately results in lower generation cost. The total costs for the said three cases of PEM elements are $543428.8, $553479.2 and $560459.4 respectively, which confirm that PEM is a measure of customer sensitivity of load shifting and higher value of PEM results in lower generation cost. The associated emissions are 257741 ton, 265248.5 ton and 266390.3 ton respectively, which show the substantial emission

Discussion on results This section involves investigation of the results obtained from different case studies. Case 1: This case is considered as the base case which is solved for economic (Wf = 1, We = 0), environmental (Wf = 0, We = 1) and combined economic-environment concern (Wf = 0.5, We = 0.5). The total cost (fuel cost + start-up cost) for these three sub cases are $565614, $576178.4 and $572912.1 and associated emissions are 269091.3 ton, 226125.3 ton and 240804.9 ton respectively. The plots of total cost and emission for these three sub cases are shown in Figs. 7 and 8 respectively, which display the trade-off between total cost and emission. The convergence characteristics for economic, environmental and combined economic-environmental solutions are shown in Figs. 9–11 respectively. Results confirm that the total cost is minimum with Wf = 1 with large emission pollution and emission is minimum with We = 1 with high total cost. The equal sharing of weighing factor Wf = We = 0.5 gives compromising results of total cost and emission compared to both the economic and the environmental solutions. Hence, further case studies are carried out for the economic solution (Wf = 1) while implementing different features of smart grid for emission reduction. The generation scheduling for the economic solution is given in Table 5. Case 2: Upcoming GVs have impressive potential of emission reduction with enhanced system reliability. Generation scheduling with randomly generated GVs and associated results are described in Table 6. Presence of GV reduces the net load demand which significantly affects the commitment of thermal units. With reduced load demand, less number of thermal units are committed and hence the total cost is reduced. The italicized units in Table 6 are the units which remain off due to the presence of the GVs compared to the base case. Hence, it is confirm that the existence of GV significantly dominates the dispatch of generating units and results in total cost of $552110.98 and emission of 263754.6 ton which are less than the base case. This total cost is also less than the reported cost of $558003 in [25]. Maximum capacity and emission plots with

Fig. 12. Maximum capacity comparison of base case with and without GV.

Fig. 13. Emission with and without GV.

Fig. 14. Load demand plot for different value of PEM.

270

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

U1

U2

U3

U4

U5

U6

U7 U8

U9

Half PEM

U10

PEM

Double PEM

Fig. 15. Scheduling of generated power for half, base and double PEM elements.

reduction compared to the base case. The cost of EDRP with double, base case and half of PEM elements are $13347.32, $7204.78 and $3746.28 respectively, which show that the EDRP

cost grows eventually as the customer’s flexibility of load shifting increases. Moreover, load demand variations of EDRP with different % of customer participation are shown in Fig. 16 which

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

indicates that the load demand reduces as the customer participation increases which consecutively reduces the total generation cost. Furthermore, as the incentive value increases, customers tend to participate more in EDRP and power demand during peak hours drops which ultimately results in lower generation cost. Load demand variation for different incentive values is given in Fig. 17. Case 4: Unpredictable wind power generation and associated (direct, overestimation and underestimation) cost is considered in this case. This study assumes a wind farm of 150 MW (100 wind power generator of 1.5 MW capacity) which is 10% of maximum load demand in addition to 10 thermal units. Cutin, rated and cut-out wind speeds of wind power generator are assumed as 3, 12.5 and 25 m/s respectively. For known value of scale factor and shape factor unpredicted wind speed can be mapped and wind power can be obtained. For parameter selection, fixing k = 2 wind power variation has been observed for different values of c from 5 to 20 which is displayed in Fig. 18. Higher value of scale factor results in higher value of wind speed which consequently generates rated power in most of the time. One of the dominating parameter in wind power

271

generator is its direct cost which varies from $2 to $4 for low wind power penetration and can be extended up to $6 for high wind power penetration level [41]. As direct cost component (d) is directly proportional to the generated wind power, increased value of d results in higher generation cost. Furthermore, selection of reserve cost (kr) and penalty cost (kp) coefficients are carried out by varying one coefficient and keeping another constant. Initially, cost analysis of wind power generation is performed assuming kp = 0 with different values of kr = 5, 10, 15 and 20 and then kr = 0 with different values of kp = 5, 10, 15 and 20 as given in Figs. 19 and 20 respectively. The individual influence of kp and kr on total wind power penetration over the scheduled time horizon (24 h) is given in Fig. 21. With the higher value of kr, wind power generation is at its rated capacity which consequently decreases thermal power scheduling and hence the generation cost. Higher value of kp results in increased penalty cost paid to the operator for not utilizing

Fig. 19. Wind power variation for different value of kr. Fig. 16. Load demand plot for different customer participation rate.

Fig. 17. Load demand plot for different incentive value.

Fig. 20. Wind power variation for different value of kp.

Fig. 18. Wind power variation for different value of scale factor.

Fig. 21. Wind power variation for combined kp, kr.

272

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

Table 7 Generation scheduling for case 3 and case 4. Hour

Generation scheduling for EDRP U1

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

455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 Total

U2

U3

U4

U5

Generation scheduling with inconsistent wind generation U6

U7

245.0 0 0 0 0 0 295.0 0 0 0 0 0 370.0 0 0 25.0 0 0 455.0 0 0 40.0 0 0 455.0 55.0 0 35.0 0 0 455.0 110.0 55.0 25.0 0 0 455.0 105.4 109.5 25.0 0 0 455.0 117.1 130.0 42.9 0 0 454.1 113.1 121.5 111.2 20.0 25.0 455.0 126.5 130.0 92.5 42.8 31.7 453.7 130.0 129.8 103.7 68.4 27.1 455.0 128.4 130.0 138.1 38.4 39.8 454.8 119.8 128.4 118.7 20.0 25.0 455.0 128.8 77.8 81.3 0 29.1 455.0 124.9 109.5 55.5 0 0 361.1 117.1 64.7 52.0 0 0 283.7 117.1 119.2 25.0 0 0 360.9 93.3 124.8 65.9 0 0 454. 130.0 130.0 30.6 0 0 455.0 128.3 130.0 83.5 28.1 25.2 437.9 120.8 0 146.6 28.0 25.0 446.8 0 0 78.1 21.0 25.4 345.8 0 0 38.5 0 0 290.6 0 0 0 0 0 cost = $553479.2, emission = 265248.5 ton

U8

U9

U10

U1

0 0 0 0 0 0 0 0 0 0 15.1 32.8 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 16.0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 455 Total

U2

U3

217.6 0 158.0 0 230.6 0 320.0 0 316.5 50.4 311.2 100.1 282.9 105.7 435.4 50.7 388.3 76.4 455.0 84.5 416.1 129.9 436.1 130.0 455.0 75.2 305.5 130.0 348.1 112.4 195.6 122.1 237.7 67.1 248.7 70.0 398.4 125.0 454.0 117.5 364.2 62.5 368.8 0 216.3 0 203.4 0 cost = $530360.8,

Fig. 22. Plot of total cost for all cases.

Fig. 23. Plot of emission for all cases.

U4

U5

U6

U7

0 0 0 0 0 0 0 0 0 25.3 0 0 0 25.0 0 0 0 28.1 0 0 52.0 31.7 0 0 67.4 89.1 0 0 83.9 25.1 0 0 121.0 52.3 30.0 26.6 129.0 40.8 20.0 44.2 93.7 108.0 50.0 25.0 98.2 127.0 20. 25.0 91.2 58.7 50.0 25.0 93.8 98.8 20.0 47.0 102.0 32.9 0 0 76.0 51.3 0 0 65.2 25.0 0 0 82.8 93.5 0 0 46.6 25.0 0 0 74.9 92.7 20.9 25.0 98.7 125.0 20.0 25.0 0 56.1 42.9 27.1 0 79.8 0 0 0 0 0 0 emission = 223813.5 ton

U8

U9

U10

PW

0 0 0 0 0 0 0 0 0 21.9 10.0 17.4 39.9 0 0 0 0 0 0 10.0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 11.7 19.6 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 21.4 0 0 0 0 0 0 0 0 0 0 0 0

29.35 137 139.1 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 148.9 141.6

273

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274 Table 8 Summarized result table. Method

Total cost ($)

Method

Total cost ($)

GABC

Total cost ($)

Emission (ton)

LR [12] GA [12] ICGA [42] BF [43] SA [14]

565,825 565,825 566,404 565,872 565,828

MA [44] GRACP [45] HAS [46] PSO-LR [47] GABC

565,827 565,825 565,827 565,869 565,614

Case Case Case Case Case Case

565,614.00 552,110.98 553,479.20 530,360.80 538,225.38 511,934.18

269,091.3 263,754.6 265,248.5 223,813.5 261,836.9 215,114.5

convergence characteristics for case 5 and case 6 are given in Figs. 24 and 25 respectively.

x 105 6.2

Total cost ($)

6 5.8 5.6 5.4 5.2

0

50

100

150

200

Number of cycles Fig. 24. Convergence characteristics of case 5.

Total cost ($)

5.8

The results are summarized in Table 8. It is clear that result gained by GABC is better than the most of the published results which proves GABC has substantial potential to solve complex, mixed integer and non-linear optimization problem of generation scheduling. It is observed that occurrence of GV enhances the system reliability by increasing total system capacity and reduces the total cost and emission compared to the case with EDRP. Moreover, test case with wind integration gives lesser results compared to the case with GV and with EDRP. But intermittent nature of wind desires another proficient substitute for cost-effective and ecofriendly solution. Consequently, collective implementation of all three DERs (GV, EDRP and wind power generator) in generation scheduling results in an economic and ecological solution.

x 105

Conclusion

0

In this paper, generation scheduling of 10 unit system is solved considering three distributed generation resources (DERs) such as distributed generation, emergency demand response program and gridable vehicles in smart grid environment. The cost of uncertainty in wind power prediction is integrated as the direct, overestimation and underestimation costs in the basic objective function of unit commitment. Incorporation of DERs in the smart grid plays a vital role to lead toward the economic and ecological solution. The well-known emergency demand response program is proved proficient to reshape the inconsistent load demand. The most crucial task of generation scheduling with stochastic wind power generation and random generation of gridable vehicle become more challenging and would become tedious and time consuming and hence Gbest artificial bee colony algorithm is executed to achieve the optimum solution. GABC proves efficient to give better solution near the optimal solution. Accumulation of all DERs in smart grid environment with the assumed intelligent infrastructure and computational techniques gives minimum total cost and emission.

5.6

5.4

5.2

5

1 2 3 4 5 6

50

100

150

200

Number of Cycles Fig. 25. Convergence characteristics of case 6.

available wind power. As kp and kr are directly proportional to the penalty and reserve cost, moderate values between 0 and 10 have been assumed with the presumption that the reserve cost of power purchasing from another source is always costlier than the penalty cost of power being not utilized. Values of d, kp and kr are purely decided by the ISO. Hence, assumed values for simulation analysis are c = 12, d = 5 with kp = 6 and kr = 8 with the presumption that the wind speed remains constant for one hour scheduling. This case results in total cost of $530360.8 and emission of 223813.5 ton which is much less than the base case with the savings of $35253.2 ($565614 $530360.8) and 45277.8 ton (269091.3–223813.5) reduction in emission. The optimal generation scheduling for case 3 and case 4 is shown in Table 7. Case 5, 6: The collective effect of GVs and EDRP results in total cost of $538225.38 and emission of 261836.9 ton. Additionally, incorporation of modern amenities of smart grid (GV, EDRP and renewables) results in total cost of $511934.18 and emission of 215114.5 ton which gives the best result among all the above discussed cases. The plots of total cost and emission for all the cases are shown in Figs. 22 and 23 respectively. The

References [1] Sousa T, Morais H, Vale Z, Faria P, Soares J. Intelligent energy resource management considering vehicle-to-grid: a simulated annealing approach. IEEE Trans Smart Grid 2012;3(1):535–42. [2] Goleijani S, Ghanbarzadeh T, Sadeghi Nikoo F, Parsa Moghaddam M. Reliability constrained unit commitment in smart grid environment. Electric Power Syst Res 2013;97:100–8. [3] Saber AY, Venayagamoorthy G. Resource scheduling under uncertainty in a smart grid with renewables and plug-in vehicles. IEEE Syst J 2012;6(1):103–9. [4] Burns RM, Gibson CA. Optimization of priority lists for a unit commitment program. IEEE/PES summer meeting; 1975. [5] Senjyu T, Shimabukuro K, Uezato K, Funabashi T. A fast technique for unit commitment problem by extended priority list. IEEE Trans Power Syst 2003;18(2):882–8. [6] Cohen AI, Yoshimura M. A branch-and-bound algorithm for unit commitment. IEEE Trans Power Appl Syst 1983:444–51. [7] Ouyang Z, Shahidehpour SM. An intelligent dynamic programming for unit commitment application. IEEE Trans Power Syst 1991;6(3):1203–9.

274

M. Govardhan, R. Roy / Electrical Power and Energy Systems 64 (2015) 260–274

[8] Daneshi H, Choobbari AL, Shahidehpour SM, Li Z. Mixed integer programming method to solve security constrained unit commitment with restricted operating zone limits. IEEE conference on EIT; 2008. p. 187–92. [9] Wang SJ, Shahidehpour SM, Kirschen DS, Mokhtari S, Irisarri GD. Short-term generation scheduling with transmission and environmental constraints using an augmented lagrangian relaxation. IEEE Trans Power Syst 1995;10(3):1294–301. [10] Ongsakul W, Petcharaks N. Unit commitment by enhanced adaptive lagrangian relaxation. IEEE Trans Power Syst 2004;19(1):620–8. [11] Sheble GB, Maifeld TT. Unit commitment by genetic algorithm and expert system. Electric Power Syst Res 1994;30(2):115–21. [12] Kazarlis SA, Bakirtzis AG, Petridis V. A genetic algorithm solution to the unit commitment problem. IEEE Trans Power Syst 1996;11(1):83–92. [13] Juste KA, Kita H, Tanaka E, Hasegawa J. An evolutionary algorithm to solve large scale unit commitment problem. In: IEEE conference on PowerCon; 2004. p. 1746–51. [14] Simopoulos DN, Kavatza SD, Vournas CD. Unit commitment by an enhanced simulated annealing algorithm. IEEE Trans Power Syst 2006;21(1):68–76. [15] Gaing Zwe-Lee. Discreat particle swarm optimization algorithm for unit commitment. In: IEEE power engineering society general meeting, vol. 1; 2003. [16] Ting TO, Rao MVC, Loo CK. A novel approach for unit commitment problem via an effective hybrid particle swarm optimization. IEEE Trans Power Syst 2006;21(1):411–8. [17] Ting TO, Rao MVC, Loo CK. Solving unit commitment problem using hybrid particle swarm optimization. J Heurist 2003;9:507–20. [18] Saber AY, Senjyu T, Urasaki N, Funabashi T. Unit commitment computation by fuzzy adaptive particle swarm optimization. IET Gener Transm Distrib 2007;1(3):456–65. [19] Mantawy AH, Abdel-Magid YL, Selim SZ. Unit commitment by tabu search. IEEE Proc Gener Transm Distrib 1998;145(1):56–64. [20] Sisworahardjo NS, El- Keib AA. Unit commitment using the ant colony search algorithm. In: Large engineering conference on power engineering; 2002. p. 2– 6. [21] Zhu G, Kwong S. Gbest-guided artificial bee colony algorithm for numerical function optimization. Appl Math Comput 2010;217(7):3166–73. [22] Jadhav HT, Roy R. Gbest guided artificial bee colony algorithm for environmental/economic dispatch considering wind power. Expert Syst Appl 2013;40:6385–99. [23] Sahebi MM, Duki EA, Kia M, Soroudi A, Ehsan M. Simultaneous emergency demand response programming and unit commitment programming in comparison with interruptible load contracts. IET Gener Transm Distrib 2012;6(7):605–11. [24] Abdollahi A, Moghaddam M, Rashidinejad M, Sheikh-El-Eslami MK. Investigation of economic and environmental-driven demand response measures incorporating UC. IEEE Trans Smart Grid 2012;3(1):12–25. [25] Saber AY, Venayagamoorthy G. Intelligent unit commitment with vehicle-togrid-a cost-emission optimization. J Power Sources 2010;195:898–911. [26] Saber AY, Venayagamoorthy G. Plug-in vehicles and renewable energy sources for cost and emission reduction. IEEE Trans Ind Electron 2011;58:1229–38. [27] Sortomme E, El-Sharkawi MA. Optimal scheduling of vehicle-to-grid energy and ancillary services. IEEE Trans Smart Grid 2012;3(1):351–9.

[28] Khodayar ME, Wu L, Li Z. Electric vehicle mobility in transmission-constrained hourly power generation scheduling. IEEE Trans Smart Grid 2013;4(2):779–88. [29] Soares J, Morais H, Sousa T, Faria P. Day-ahead resource scheduling including demand response for electric vehicles. IEEE Trans Smart Grid 2013;4(1):596–605. [30] Sousa T, Morais H, Soares J, Vale Z. Day-ahead resource scheduling in smart grids considering vehicle-to-grid and network constraints. Appl Energy 2012;96:183–93. [31] Hetzer J, Yu DC, Bhattarai K. An economic dispatch model incorporating wind power. IEEE Trans Energy Convers 2008;23(2):603–11. [32] Yusta JM, Khodr HM, Urdaneta AJ. Optimal pricing of default customers in electrical distribution systems: effect behavior performance of demand response models. Electric Power Syst Res 2007;77:548–58. [33] Albadi MH, EL-Saadany EF. A summary of demand response in electricity markets. Electric Power Syst Res 2008;78(11):1989–96. [34] Aalami HA, Moghaddam MP, Yousefi GR. Demand response modeling considering interruptible/curtailable loads and capacity market programs. Appl Energy 2010;87(1):243–50. [35] Aalami HA, Moghaddam MP, Yousefi GR. Modeling and prioritizing demand response programs in power markets. Electric Power Syst Res 2010;80(4):426–35. [36] Seguro JV, Lambert TW. Modern estimation of the parameters of the weibull wind speed distribution for wind energy analysis. Wind Eng Ind Aerodyn 2000;85(1):75–84. [37] Chatterjee A, Ghoshal SP, Mukherjee V. Solution of combined economic and emission dispatch problems of power systems by an opposition-based harmony search algorithm. Electrical Power Energy Syst 2012;39:9–20. [38] Basturk B, Karaboga D. An artificial bee colony (ABC) algorithm for numeric function optimization. In: IEEE swarm intelligent symposium; 2006. p. 12–4. [39] Swarup KS, Yamashiro S. Unit commitment solution methodology using genetic algorithm. IEEE Trans Power Syst 2002;17(1):87–91. [40] Chandram K, Subrahmanyam N, Sydulu M. New approach with muller method for profit based unit commitment. In: Power and energy society general meeting- conversion and delivery of electrical energy in 21st century; 2008. p. 1–8. [41] Albadi MH, El-Saadany EF. Overview of wind power intermittency impacts on power systems. Electric Power Syst Res 2010;80(6):627–32. [42] Damousis IG, Bakirtzis AG, Dokopoulos PS. A solution to the unit commitment problem using integer coded genetic algorithm. IEEE Trans Power Syst 2004;19(2). [43] Eslamian M, Hosseinian SH, Vahidi B. Bacterial foraging-based solution to the unit-commitment problem. IEEE Trans Power Syst 2009;24(3). [44] Jorge V, Smith A. A seeded memetic algorithm for large unit commitment problems. J Heurist 2002;8:173–95. [45] Viana A, de Sous JP, Matos MA. Fast solutions for UC problems by a new metaheurisic approach. Electric Power Syst Res 2008;78(8):1385–95. [46] Pourjamal Y, Ravadanegh SN. HAS based solution to the UC problem. Electric Power Energy Syst 2013;46:211–20. [47] Balci HH, Valenzuela JF. Scheduling electric power generators using particle swarm optimization combined with lagrangian relaxation method. J Appl Math Comp Sci 2004;14(3):411–21.