Internet of medical things-load optimization of power flow based on hybrid enhanced grey wolf optimization and dragonfly algorithm

Future Generation Computer Systems 98 (2019) 319–330

Contents lists available at ScienceDirect

Future Generation Computer Systems journal homepage: www.elsevier.com/locate/fgcs

Internet of medical things-load optimization of power flow based on hybrid enhanced grey wolf optimization and dragonfly algorithm ∗

Shilaja C. , Arunprasath T. Department of Electrical & Electronics Engineering, Kalasalingam Academy of Research and Education, Krishnankoil, India

highlights • This research presented a hybrid algorithm based on Enhance grey wolf optimisation and algorithm of Dragonfly for handling OPF (optimal power flow).

• The efficiency of the system is experimented on IEEE30 bus system and verified using wind energy system. • The presented method is fast and effective and it is evidenced by comparing the presented method with conventional method. • The presented hybrid method provides various benefits from EHO as well as GWO methods cognitive load which suitable to resolve major OPF issues.

article

info

Article history: Received 27 October 2018 Received in revised form 11 December 2018 Accepted 29 December 2018 Available online 29 March 2019 Keywords: Hybrid renewable energy system Enhance grey wolf optimization Dragonfly algorithm Optimal power flow IEEE30 bus system

a b s t r a c t In hybrid renewable energy system, optimization and control are more complex and non-linearity in nature. This research presented a hybrid algorithm based on Enhance grey wolf optimization and algorithm of Dragonfly algorithm for handling OPF (optimal power flow) issues. The hybrid algorithm is proposed for solving the minimization of fuel cost, power loss and voltage deviation. The Renewable energy are incorporated with solar and wind energy. To forecast the wind and solar photovoltaic power output the weibull distribution function are modelled. The traditional method is slow and incapable to solve non-linearity problems. The proposed method is fast and effective and it is experimented on IEEE 30 bus system by comparing with recent existing methods. © 2019 Elsevier B.V. All rights reserved.

1. Introduction Modern techniques and usage of power in a wide range in corporate bodies bring the necessity in the distribution of electricity in most of the countries. This issue leads to a tremendous increase in the production of electricity. Sequentially, the power management in a rapid level of engineering and technology is represented as a major investment in economics and commodity. This enhancement in the production results in the perception of high quality grids and renewable energy system. The foremost objective of high quality grids and electrical systems is creating less expensive and a secured system. In this point, electrical system needs a proper scheduling and high performance that is worthy to most of the researchers. OPF is the tool to achieve this major goal that is rapidly efficient and effective in performing the power systems. OPF method is used to execute RES to secure and reduce the cost and difficulty. The optimal power flow method is mainly used to find out the best solution for reduced cost and ∗ Corresponding author. E-mail address: [email protected] (Shilaja C.). https://doi.org/10.1016/j.future.2018.12.070 0167-739X/© 2019 Elsevier B.V. All rights reserved.

improved power efficiency. The critical objective for OPF is to use the power system on efficient and reduced cost. The optimum solution obtained in OPF represents the optimization of the cost in operating in electrical systems. This OPF solutions approaches are divided into predictable and artificial intelligence method. The major objective in this issue is voltage deviation reduction, energy loss and enhancing the stability of the voltage. In this point, reactive energy compensation sources such as reactors, voltage generator bus, capacitor and on-load tap-chargers are used. Indeed, there may be many huddles such as bus voltage, flow of load and reactive energy output in generators which has to be met. Even though, energy output of generators, continuous variables in voltage which are normally distinct when compare to transformers, capacitors of shunt and reactor output. Number arithmetic process is engaged to resolve these issues. Optimizations of gradient methods are involved in [1–4]. This kind of interior method is used to address ORPD issues [5]. To optimize a system, second order schedule is utilized [6]. The most of the mentioned methods have many disadvantages and limits in nonlinear issues. Probably, there may be distinct huddles, processes and issues with various basic points. A universal optimum

320

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

2. Literature review

Nomenclature Abbreviations OPF ICA PSO GA OSAMGSA GWO FACTS MSA ABC DA APSO PV ISO PDF

optimal power flow imperialist competitive algorithm particle swarm optimization genetic algorithm opposition-based self-adaptive modified gravitational search algorithm grey wolf optimizer flexible AC Transmission system moth swarm algorithm artificial bee colony dragonfly algorithm aging particle swarm optimization photovoltaic independent system operator probability density function

Symbols NTG min PTGi h

k

P ss,k g j P w s, KRW ,j

f

Pwav,j fw (pw,j ) Kpw,j Pwτ ,j Psav,k G2 (u ,v) f i

Overall count of thermal generators minimal energy in all the thermal unit that generates the operation cost coefficient with kth solar power plant power from the plant cost coefficient of jth wind energy plant energy from the plant coefficient of reserve cost that consist of jth wind energy plant real offered energy from the plant wind energy probable quantity process of jth wind energy plant cost of the penalty coefficient to the jth wind systems output energy from the wind system real obtainable energy from the system total deviation profile ith generator fuel value

computation is lost while obtaining the capability. The above mention methods went through basic overview to get rid of these margins. In remote areas the renewable energy usually available by wind and solar energies. Yet, by the variation of periodical and seasonal reasons the energy supply can practically change as either separate wind energy or solar energy. To meet the certainty demands the multiple energy resources has enhanced simultaneously. This idea is not effective in resolving the quantitative reactive energy dispatch in large systems. Previously, various probable researches have been made to resolve this optimization issues. ICA (imperialist competitive algorithm), PSO (particle swarm optimization) and GA (Genetic algorithm), are few among them [7–15]. An ORPD issue has been resolved using new method proposed in [16]. Its major merit is optimizing distinct system and sequential variables in decisions. In OSAMGSA (oppositionbased self-adaptive modified gravitational search algorithm) is represented in cracking various objective ORPD issues.

In [17] was presented nature GWO (Grey Wolf Optimizer) to the computation of OPF issues. The fundamental IEEE 30 bus systems were introduced to assess the efficiency and excellence representing the above algorithm. This work measured couple of various issues such as cost of the fuel and deviation in the voltage to estimate the quality of STATCOM. Overall analysis of the result clearly states that STATCOM usage provides the better results in both the issues. Presented algorithm of GWO provides the most favourable results in control variables when compared with other wide-ranging algorithm that was resulted in the studies such as ACO and PSO. While utilizing various FACTS (flexible AC Transmission system) such as UPFC, TCSC, and SVC, enhances the energy transformation of strength and capacity in the system. In paper [18] were provides unique MSA (Moth Swarm Algorithm), influenced by the moths orientation directed to the moonlight to resolve distinct OPF issues. To enhance the ability of exploration and exploitation, the collaborative mechanism of learning with immense memory capacity and diversity in population across the levy mutation has been proposed along with spiral motion and Gaussian walks with adaptive method. IEEE 30 bus, IEEE 57 bus and IEEE 118 bus carries algorithms of heuristic search and MSA systems. Various energy system holds basic energy generation, ratio of load tap changer, capacitance value of shunt, voltage of bus and basic energy generation has been applied and optimized through these approaches. In [3,5,9–11] were presented method takes more computational time. This presented method has experimented on IEEE-30 buses structure a method in Monarch Butterfly Optimization towards resolve the OPF issues. In paper [19] showed ABC (Artificial Bee Colony) algorithm that resolveOPF issue. This method was implemented in IEEE 14 buses as well as IEEE 30 buses trial systems. This method is slightly faster than PSO and GA. In paper [2] were framed an optimization method in a hybrid meta-heuristic, DA (Dragonfly Algorithm) and APSO (Aging Particle Swarm Optimization) to handle OPF issues. The above mentioned algorithm is applied to obtain the power system control variables optimal values and to rectify OPF problem. The minimization of power loss, deviation in voltage profile and fuel cost control is the main purposes of OPF issues. The efficiency of intended hybrid algorithm is practised in IEEE-30 buses. The obtained result is measured by power loss constraints using the wind energy system. In [1] has resolved the problematic of OPF in a design in corporate through various form in wind aiming the goal of minimizing the power rate production and the energy grid loss minimization by using GWO algorithm. An advanced edition of IEEE-30 buses trail system is applied based on the issues of OPF in the proposed method. GA is compared with obtained results from the study. The efficiency of the method proposed in terms of both dynamic speed and the ultimate results that depicts the originality of the method. In [4,8] proposed search gravitational algorithm using a unique hybrid PSO to solve environment/economic with most favourable energy flow delivering integrated wind energy generation. The highest probability in wind energy generation projects through computation the speed of the wind in highest power. The quantitative results that obtained from this designed algorithm depicts that PSO-GSA algorithm can provide greater accuracy to resolve a true quality energy that flows with the rapid convergences. In [6] presented a novel enhanced PSO technique to eliminate immature converges when optimizing the whole expenses of the power in the proposed system. This system includes battery bank, wind turbine, diesel generator and Photovoltaic Generator. The

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

321

obtained result from the method proposed is compared with iterative technique. In [20] it is estimated in accordance to the source of the renewable energies like wave, wind, biomass, solar and geothermal. A complete study of the current techniques and sources are determined in these researches and illustrated here. The obtained result tells that the solar and the wind are the energies which is most appropriate sources of energy. This paper also proposes a novel model of solar and wind hybrid power system.

In certain cases, the real power produced from the wind power plant may be lesser than the estimated quantity. Generally, this is known as over estimation of energy from the less certainty sources. To deliver continuous power supply to the customers the operation system needs to use results from the spinning scenarios. Reserve cost is said to be the cost that commits the reserve producing systems to meet these overestimated quantity. The below reserve cost equation of the jth wind energy plant is defined as

3. Mathematical model

CRW ,j (Pws,j − Pwav,j ) = KRW ,j (Pws,j − Pwav,j )

Adapted IEEE 30 bus system includes solar PV, wind generators and thermal. Solar and wind PV energy outputs are known to be variables between each other. The changes in the output of the energy should be equalized with output of the generator and the reserve combination. Overall generation rate includes the cost of the operation in all generators, penalty rate and reserve rate that are described in the sequence. a. Thermal Power Generators in the Cost Model The fossil fuel is the most essential resource in all thermal generator units. The approximate value of the quadratic relation is associated between fuel rate and the power generated. CT 0(PTG ) =

NTG ∑

2 ai + bi PTGi + ci PTGi

(1)

(5)

Pws,j

∫ KRW ,j

(Pws,j − Pw,j )fW (Pw,j )dpw,j

(6)

0

Where KRW ,j represents the coefficient of reserve cost that consist of jth wind energy plant, Pwav,j represents real offered energy from the plant. fw (pw,j ) Represents wind energy probable quantity process of jth wind energy plant. In contrast to the case of over estimation, there may be a situation where the actual wind produced from the system may be greater than that of the estimated energy. In this case, the power generating sources are under estimated. Abundant energy may be wasted if it is not utilized properly by the traditional generators. ISO must take responsibility for the cost and the wastage of this enormous amount of energy. Cost taken as penalty from the jth wind systems can be defined as follows,

i=1

Where ai , bi , ci represents coefficients of the cost in ith thermal generator providing energy output PTGi . Overall count of thermal generators is NTG . The most original and the précised model with cost control should be considered in the effect of valve point. Greater variables in fuel rate functions exhibits a multi valve turbines of steam is generated by thermal units [3]. A sinusoidal process of multi valve steam turbines is designed using valve loading effect, here with the values taken to the fundamental rate function as shown in Eq. (1). Overall generation rate is represented as follows. CT (PTG ) =

NTG ∑

2 min ai + bi PTGi + ci PTGi + |di ∗ sin(ei ∗ (PTGi − PTGi ))|

(2)

i=1

Thus ei and di represents the coefficients of valve point loading min is the minimal energy in all the thermal unit that effect. PTGi generates the operation. b. Direct wind cost and photovoltaic solar energy Fuel is not required for wing energy generators and solar PV like other traditional thermal energy generator. ISO plays a major role in assigning few pay back cost for the initial output of solar/wind PV plants or for the maintenance and cost for remuneration when the solar/wind power plants owned [19]. But when private parties own solar/wind PV plants then ISO must pay some price to schedule the energy in the contract bases. The total cost used for generating wind energy with the help of jth plan is designed as a process for the planned energy power. Cw,j (Pws,j ) = gj Pws,j

Cpw,j (Pwav,j − Pws,j ) = Kpw,j (Pwav,j − Pws,j )

∫

Pwτ ,j

(pw,j − pws,j )fW (pw,j )dpw,j

= Kpw,j

(7)

Pws,j

Here, Kpw,j represents the cost of the penalty coefficient to the jth wind systems, Pwτ ,j represents the output energy from the wind system. d. Evaluation of Cost in solar power uncertainties As wind systems, solar plant system consists of less certainty output and irregular supplies. Basically, situations like under estimating the power generation and over estimating the power generation may happen in solar systems also. But, the solar rays constrains lognormal values PDF, varies from wind supplies that is known for following Weibull PDF, to make the calculation easy in terms of reserve and penalty cost to build the basic concept represented in [8,20]. Cost reserve in kth solar systems are as follows, CRS ,k (Pss,k − Psav,k ) = KRS ,k (Pss,k − Psav,k )

= KRs,k ∗ fs (Psav,k < Pss,k )∗ [Pss,k − E(Psav,k < Pss,k )]

(8)

Where KRs,k a reserve represents cost coefficient consists of kth solar system. Psav,k Represent real obtainable energy from the system. fs (Psav,k < Pss,k ) Represent probable of shortage in solar energy that occurs from the designed power plant. Pss,k − E(Psav,k < Pss,k ) is denotes the available solar power system. Pss,k Represents the cost of penalty in under estimated in kth solar system is follows,

(3)

Here, gj represents cost coefficient of jth wind energy plan, Pws,f represents the designed energy from the plant. Like wind energy plant, the total cost depicting to kth solar energy plant is,

CPs,k (Psav,k − Pss,k ) = KPs,k (Psav,k − Pss,k )

= KPs,k ∗ fs (Psav,k > Pss,k )∗ [E(Psav,k > Pss,k ) − Pss,k ]

(9)

Here, hk represents cost coefficient with kth solar power plant. Pss,k represents designed power from the plant.

Here KPs,k represents penalty rate coefficient pertaining to kth solar power systems, fs (Psav,k > Pss,k ) is probable to solar energy output compare to the designed power (Psav,k > Pss,k ) − Pss,k denotes the estimated solar power Pss,k

c. Evaluation of Cost in wind power uncertainties

e. Solar and Wind PV power model

Cs,k (Pss,k ) = hk Pss,k

(4)

322

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

In the scheduled power houses, the wind power connects to bus 5 is similar to the output energy of 25 turbines when the output wind form that has 20 turbines which connect to bus 11. All the individual power turbine which are rated as 3MW corresponds to the real output energy from the wind turbines depending on the speed of the wind which are encountered. f. Estimation of probability in wind power In regards to Eq. (31) it is said to be that variables in wind energy is distinct in both regions corresponding to the speed of the wind. If wind power (v) is under the cut in speed (vin ) more than cut out speed (vout ) then the output of the power is zero. The power output by turbine Pwt among the estimated wind speed (vr ) and cut out speed (vout ). In this distinct zone the probable values are shown below, fw (pw ){pw = 0} = 1 − exp[−( fw (pw ){pw = pwr } = 1 − exp[−(

vin c

vr c

)k ] + exp[−(

)k ] + exp[−(

vout c

vout c

)k ]

(10)

)k ]

(11)

The output of wind turbine is spontaneous among cut in speed vin and rated speed vr of the wind. The probable value of the spontaneous region is given below, fw (pw ) =

k(vr − vin )

pw (vr − vin )]k−1 pw r vin + ppwwr (vr − vin ) k ) ] × exp[−( c c k ∗ pw r

|[vin +

fw (pw ){pw = pwr } = 1 − exp[−(

vr c

)k ] + exp[−(

(12)

vout c

)k ]

(13)

∑

∗

[Pss − Psn− ] fsn−

(14)

n=1

Where Psn− represents the power which is available lesser than that of the left half plan as Pss , fsn− denotes the particular frequency in occurrence Psn− N- denotes the number of distinct bins in left-half Pss can also be represented as pairs. Pss − Psav for the PDF. The accuracy of the obtained results can be improved to a limit when the number of segment increased. Like wise cost of under estimation value in Eq. (8) can be estimated as, N+ ∑

[Psn+ − Pss ]∗ fsn+

(15)

n=1

Where Psn+ denotes the power availability greater than the power scheduled, Pss , indicates the right-half plane Pss from the shown histogram. fsn+ Denotes the similar frequency occurred Psn+ . N+ denotes the count of distinct bins in right half Pss can also be denoted as (Psn+ , fsn+ ) generated for the PDF.

In power system, the control variables are a complex and non-linearity problem. Hence, OPF is the technique to solve this complexity and non-linearity and obtain the optimal solution for cost and power loss. To minimize f(m,n)

{

E(m, n) IE1 ≤ IE(m, n) ≤ IEU

G1 (u, v ) =

NG ∑

bfi ($/h)

(17)

i−1

Here bfi represents the cost fuel generator i. 2 ($/h) bfi = ki + li PGi + bi PGi

(18) th

Where ki , li and bi be the rate coefficients of generator i . iii. Deviation of Voltage profile The principal objective of voltage profile deviation is to reduce total voltage deviation PQ bus. The cost of fuel optimization and voltage profile development are collectively called two-fold function, and it is expressed as: G2 (u, v ) =

NG ∑

fi ($/h) + V1

∑

| P . U − Vi |

(19)

i∈NL

where, G2 (u,v) denoted as total deviation profile fi is the ith generator fuel value V1 is a user-defined weighing factor. iv. Minimization of Power loss Pl denotes Total loss of the power is estimated below, N1 ∑

rk

r2 k−1 k

+ x2k

[Vi2 + Vj2 − 2Vi Vj cos(θi − θj )]

(20)

N1 represents the transmission lines count, rk is a value of confrontation and xk is value among the line jth and ith bus, Vi , Vj denotes magnitudes of the voltage and θi θj denotes the angles of the voltage in the bus j and i. When wind energy systems are without constraints the fundamental work is defined as fi (di ) that reduces the function cost in EGWO–DA method. Most of the wind energy plants are practised in rural areas of power demands. Hence, the presented method is experimented in Wind power plant to minimize the cost and power loss. f (d) =

Nd ∑

fi (di ) + B + T

(21)

i

Here T and B denotes the coefficients which are lost in the generator 5. Proposed methodology 5.1. EGWO (Enhanced Grey Wolf Optimization algorithm)

4. Problem formulation

Subjected to

ii. F Mitigation of fuel cost The mitigation of fuel cost is expressed as,

Pl =

N−

Cps (Psav − Pss ) = KPs (Psav − Pss ) = KPs

i. Processes for intention The principal intention process sis for resolve the OPF issue is provided in follows.

i−1

g. Estimation of solar energy under/over cost evaluation The dotted line denotes the scheduled energy from the solar systems which actually delivers the grid. The designed power would be any quantity of power which are mutually accepted among solar power plant owners and ISO which is already quoted before. CRs (Pss − Psav ) = KRs (Pss − Psav ) = KRs

Here f (m,n) represented as objective process system, IE(m,n) represents inequality constraints, E(m, n) represented sameness limitations of the ith generator, and IEU is a higher level of inequality constraints and IE1 is ith generator’s lesser point of dissimilarity limitations.

} (16)

EGWO algorithm is developed by the inspired behaviour of hunting and social nature of wolves. To extend the algorithm, the behaviour of wolves is mathematically, and the leader denotes alpha (α ) wolf. Sequentially, corresponding to good results named beta (β ) and delta (δ )wolves. The remaining obtained results can be estimated as omega (ω) wolves. The technique of optimization is followed by alpha, beta and delta in GWO method. The omega wolves come next to the alpha, beta and delta wolves to

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

search the universal optimization. The below equation express the behaviour of the hunting wolves.

− →

⏐− → ⏐→ −

− →

⏐ ⏐

D = ⏐ C . XP (t) − X (t)⏐

− →

− →

(22)

− →− →

X (t + 1) = Xp (t) − A . D

− →

− →

Here t denotes the number of iteration, A and C are a − → − → coefficient vector. X p Is the place of victim X is a place of the wolves.

− → − →− → − → A = 2 a . r1 − a − → − → C = 2. r2

(23)

Here elements in continuously decrease the value 2 to 0 above the progression in iterations. Where r1 , r2 indicates unsystematic vectors [0,1]. The pursue and leadership manners of grey wolves are given below:

− →

⏐− → ⏐→ −

− →⏐⏐

(24)

− →

⏐− → ⏐→ −

− →⏐⏐

(25)

Dα = ⏐ C1 . Xα − X ⏐

D β = ⏐ C 2 . Xβ − X ⏐

behaviours of dragonfly. Both the swarming behaviours look comparable in both the two key stages in optimization through meta heuristic algorithms, namely, investigation and utilization. In the swarm of static, all the dragonflies form sub-swarms to take off over the various areas, that is the main motive of investigation stage. A. Modelling of Artificial Dragonflies (1) The behaviour of dragonflies can be summarized in terms of five steps, namely, departure, arrangement, consistency, magnetism in the direction of a food source and Distraction towards the outside an enemy [21]. The departure is estimated as follows: si = −

N ∑

⏐− → − →⏐⏐ ⏐→ − Dδ = ⏐ C3 . Xδ − X ⏐ − → − → − → − → X1 = Xα − A1 .( Dα ) − → − → − → − → X2 = Xβ − A2 .(Dβ ) − → − → − → − → X3 = Xδ − A3 .( Dδ ) − → − → − → X1 + X2 + X3 − → X (t + 1) =

(26) (27) (28) (29) (30)

3

These c components are used to check the optimal local results. EGWO method shows three good results that represents alpha, beta and delta. These are denoted as leaders. To identify the best solution close to the universal optimization these mentioned leaders helps to search the agents to reach the designated regions in the search space. In proposed EGWO algorithm different advancements are done so as to find a solution to problems of efficiency and convergence rate. The →parameter is considered as a a

random vector in the range [0, 1] whose values are crucial in providing balance between exploration and exploitation. Adaptive values of → parameter maintain exploration to prevent getting a

trapped in local optima and cope up the accuracy problem. It is critical parameter in redefining the resulted vectors and efficiently utilized in operating the rate convergence in the method. With this intention that simulates the behaviour of hunting grey wolves an assumption is made that the fittest wolf α has better knowledge on the efficient place of the prey, and for this reason only the fittest solution is saved. The global best solution obtained from the entire population helps in achieving the global optimum. The EGWO algorithm is proposed in order to provide enhanced performance in terms of avoiding getting trapped in pre-mature convergence, convergence rate and accuracy. 5.2. Dragonfly algorithm Dragonfly is supposed to be a tiny predator in nature, which chases mostly the other little insects. An appealing truth on a dragonfly is its distinctive and extraordinary swarming behaviour. The main purpose of the swarm of Dragonflies is used to hunt and migrate. The former reason is known static or feeding swarm, the other one is known as the migratory or dynamic swarm [21]. The foremost motivation of Dragonfly algorithm begins in static as the initial point and the energetic swarming

X − Xj

(31)

j=1

Here X shows place of particular individual, Xj is position of jth near by individual, N is count of the nearby individuals. (2) The position is computed below:

∑N

− →

323

Ai =

j=1

Vj

N Here, Vj shows the velocity of jth nearby individual. (3) Consistency is computed as:

∑N ci =

j=1

Xj

N

−X

(32)

(33)

(4) Attraction towards food source Fi = X + − X

(34)

Here, X and X+ show place of present individual corresponding to the source of food, respectively (5) Interruption noticeable of enemy is computed below: Ei = X − + X

(35)

−

Here, X and X are place of the existing individual and an enemy, respectively Ei = X − + X For updating place of the simulated dragonfly in the research space to simulate the progresses of dragonflies, both the vectors, namely step (∆X) vector with position (X) vector are considered. The stage vector used in DA is similar to the vector velocity of PSO. Actually, the Dragonfly Algorithm is developed using the framework of Particle Swarm Optimization technique. In dragonfly method, the stage vector ∆X provides the direction of progress of the dragonflies. The step vector can be calculated as given below:

|∆Xf +1 = (sSi + aAi + cCi + fFi + eEi ) + w∆Xf

(36)

Here s indicates parting mass, Si indicates parting of ith entity, a indicates configuration mass, Ai indicates ith individual configuration, c is consistency mass, Ci indicates ith individual consistency, f show food aspect, Fi indicates food resource in ith entity, e indicates enemy source, Ei shows position of the enemy of ith entity, w indicates indolence weight, t shows iteration count.

|∆Xf +1 = (sSi + aAi + cCi + fFi + eEi ) + w ∆Xf Xf +1 = Xf + Lev y(d)∗ Xf

(37)

Once step vector ∆X has been computed, the next step is to compute position vectors Xas : Xf +1 = Xf + ∆Xf +1

(38)

Here t shows current iteration. With the separation factor s, the alignment factor a, the cohesion factor c, the food factor f,

324

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

and the enemy factor e, various extraordinary with behaviours can be attained throughout the optimize process. While there is no nearby solution, the place of dragonflies is modernized by means of a random walk (Levy flight). Thus, the place vectors X are estimated as: Xf +1 = Xf + Lev y(d)∗ Xf

(39)

The neighbourhood area is increased and ultimately, at final phase of the optimization process, the swarm become only one group. Food resources with the opponent are selected from good along with the poor results obtained in the every swarm at any instant. This leads the convergence towards the promising areas of research space and at the similar time, it leads divergence outward the non-promising areas in search space.

utilizing various sets of uniform number distributing ranges of the boundaries. xi = xmin + u(xmax ) − xmin i i i

(40)

Here xi denotes rate of the ith wolf consequent to the ith direct variable, xmax along with xmin are the higher and lesser limits of i i the ith direct variablealong with u denotes a identical chance in the time [0,1]. Also, create a repository for pareto-set. By utilizing step 5 to estimate the fundamental process in the course of iteration most of the wolves perform the whole NR (Newton–Raphson) analysis. By way of step 6, the basis of the obtained result from the NR power flow the value of the fundamental process is calculated. The method of Paretois

∀i = {1, 2, . . ., n}, Fi (x1 ) ≤ F i(x2 )

(41)

e j = {1, 2, . . ., n}, Fj (x1 ) ≤ Fj (x2 )

(42)

5.3. Hybrid EGWO- DA GWO method is based on the SI technique of optimization that resembles the leadership of grey wolves in term of its behaviour in hunting. This technique consists of two adjustments in basic GWO. Primarily, increase in the diversity with the opponent learning technique basis. Then the parameter value that oscillates between the value [2, 0] for closely 75% iteration and obtains few sustained values for the pending period. The DA gives investigation and utilization phase. When combining these two phase of Dragonfly method that targets ultimate results which could be obtained. Steps were involved in Hybrid EGWO–DA in below. Initializing according to the step 3, the following equation shows the initial search agents that take up the place frequently

Stored in the primary set of Pareto is then obtained after the application. Comparison of strategy in dominance is adapted to update the archive is as follows,

• It is important that, they obtained results may not be saved in the repository if the obtained value is weak or subjugated by the neighbouring number in the colony. • The obtained results can be saved in the repository when the non dominated neighbour may not be dominated by any results in the existing depositary.

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

325

Fig. 1. Enhanced Grey Wolf Optimization algorithm flowchart.

• All the other dominated result in the existing depositary will be eliminated by the non dominated individual. Through step 7, inserting the existing non dominated individual in the depositary is done. All the dominated places are rejected in the process by the depositary. The below equation followed by step 9, A = 2a.r1 − a

(43)

C = 2.r2

(44)

Here a decreased sequentially from 2 to 0, r1 and r2 are distributed randomly within the range [0, 1]. In step 10 the boundary limits verification has certain case, destruction in the dissimilarity limits, the place of the corresponding wolves should be fixing by utilizing xLim =

{

xmax xmin

if x > xmax if x < xmin

(45)

Step 14 should be reached instantly when most number of iteration are satisfied, if not step back to step 4. As soon as the preys stop, the chasing process of the grey wolves will also be stop. From the repository the ultimate pareto set is obtained.

6. Simulation results The proposed Enhanced Grey wolf Optimization–Dragonfly method is applied for solving OPF issue in the model IEEE 30- bus analysis system [17, 30]. This test system consists of 6 originators at buses 1, 2, 5, 8, 11 and 13. This system has 4 number with off nominal tap ratios and transformers at line nos. 6 to 9, 6 to 10, 4 to 12 and 28 to 27. Jolt VAR compensation has been provided on bus nos. 10, 12, 15, 17, 20, 21, 23, 24 and 29. The system has all system require of 2.834 pu for real energy, and 1.262 pu for the reactive energy at base of 100MVA. Bus no. 1 has been considered as slack bus. The generator (PV) bus voltage magnitude limits were considered as 0.95 pu and 1.1 pu, respectively while at load buses, the minimum and maximum voltage magnitudes were assumed as 0.95 pu and 1.05 pu, respectively. As many as, 20 trials were taken using Enhanced Grey wolf Optimization– Dragonfly method for solving the OPF issue and the best results are given here IEEE 30 bus system information obtained from the data set using the literature review is shown in Table 1. The hybrid algorithm is used to find the direct variables with optimize the values. In IEEE 30 bus system the hybrid method is given to optimize

326

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

Fig. 2. Dragonfly algorithm flowchart. Table 1 IEEE 30 bus system data.

Table 2 Different cases studies investigated in this paper.

IEEE 30 Characteristics

Value

Details

Buses

30

[11]

Branches

41

[11]

Generators

6

Buses: 1, 2, 5, 8, 11 and 13.

Load voltage limits

24

[0.95–1.05]

Shunts VAR compensation

9

Buses: 10, 12, 15, 17, 20, 21, 23, 24 and 29.

Transformers with off-nominal tap ratio

4

Branches: 11, 12, 15 and 36.

Control variables

24

–

Name

Objective function to be minimized

Constraints

Test system

Case 1

Quadratic fuel cost

Equality and non-equality

IEEE 30

Case 2

Active power loss

Equality and non-equality

IEEE 30

Case 3

Quadratic fuel cost considering voltage deviation

Equality and non-equality

IEEE 30

This paper intended to decline the dynamic energy loss for every programme line by minimizing the equation,

the solution. The presented method of fuel cost is compared with

f =

Case 1: Minimization of fuel cost The main goal of minimizing the overall cost of fuel in energy generation is the first major function, and it is expressed as follows,

f =

) 2 ai PGi

+ bi PGi + Ci + penality ($/h)

(46)

i=1

Here, ai , bi and Ci denotes the cost coefficient of the ith generator. In this phase, the DA–EGWO generates a good solution when it is linked to the APSO and ABC techniques. Case 2: Minimization of power loss

Gij Vi2 + 2Vi Vjcos (δi − δj ) + Penality (MW)

(47)

i=1 j̸ =i

other algorithm is shown in Table 2.

( NG ∑

nl nl ∑ ∑

Case 3: Minimization of fuel cost and voltage deviation The profile of the voltage is said to be the major key for system amenity quality. The enhancement of the voltage profile consists of decreased aberration in the load bus voltage to the unit. Remember, the cost basis of the basic function that indicates to a viable results with detrimental voltage deviation. Hence, a various goal is taken sequentially to reduce the cost of the fuel along with a voltage deviations (VD), that denotes the equation given below f =

( NG ∑ i=1

) 2 ai PGi

+ bi PGi + ci + ℓVD

NL ∑

|VLi − 1| + Penality

(48)

i=1

The comparison between the presented method EGWO–DA, Improved GA, PSO, and DA-APSO are given in Table 2. The Table 2 proves that the presented method is improved than the other

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

327

Fig. 3. Hybrid EGWO–DA flowchart. Table 3 Fuel rate comparison between different algorithms. Methods

Fuel cost ($/h)

EGWO–DA DA-APSO [2] IMPROVED GA [6] APSO [2] ABC [19]

790.56 802.63 805.80 803.65 803.78

method. The cost is reduced, and power loss is minimized (see Table 3). The voltage deviation, fuel cost and two-fold objective process are optimized in the presented algorithm. Optimized results are analysed with other algorithm and are tabulated in Tables 4 and 5. 6.1. Minimization of fuel cost Generally, all the six generators is considered in quadratic form for cost function. Approximate cost of fuel is estimated as 800.6594 $/h. The Fig. 1 shows the cost of the fuel EGWO– DA and various current techniques in EGWO–DA offers the best minimization of the fuel cost.

Table 4 Comparison between presented algorithm and other algorithm. Control variables

IMPROVED GA [6]

PSO [4]

DA-APSO [2]

EGWO–DA

PG1 (MW) PG2 (MW) PG5 (MW) PG8 (MW) PG11 (MW) PG13 (MW) V1 (p.u.) V2 (p.u.) V5 (p.u.) V8 (p.u.) V11 (p.u.) V13 (p.u.) T11 (p.u.) T12 (p.u.) T15 (p.u.) T36 (p.u.) QC10 (MVar) QC12 (MVar) QC15 (MVar) QC17 (MVar) QC20 (MVar) QC21 (MVar) QC23 (MVar) QC24 (MVar) QC29 (MVar)

1.775094 0.48722 0.21454 0.20954 0.11768 0.12052 0.081 1.063 1.034 1.038 1.100 1.055 1.000 0.975 0.975 1.000 0.001 0.007 0.019 0.024 0.015 0.022 0.047 0.047 0.024

1.7696 0.4783 0.2120 0.2178 0.1192 0.1092 1.0937 1.0703 1.0692 1.0302 1.0293 1.0116 1.0982 0.0452 0.0122 0.0456 0.0356 0.0498 0.0145 0.0582 0.0489 0.0239 0.0111 0.032 0.052

1.7453 0.04873 0.2027 0.2110 0.1235 0.1133 1.023 1.023 1.037 0.0912 1.0092 0.982 0.9112 0.0112 0.0311 0.0342 0.0124 0.0311 0.0300 0.0101 0.067 0.051 0.032 0.0144 0.012

1.7395 0.4871 0.2095 0.2098 0.1201 0.1192 1.032 1.004 1.031 1.002 0.998 0.996 0.9020 0.0356 0.0256 0.0262 0.0122 0.0299 0.0226 0.0126 0.021 0.025 0.023 0.0156 0.018

Fuel cost ($/h)

805.805

805.65

802.63

790.56

6.2. Minimization of power loss High total active loss is obtained from the mean total of DA method with 3.1455 MW. Fig. 2 demonstrate the minimization of the power loss in DA–EGWO and various other methods. This proposed method reduces the active power loss to 2.003 MW. (see Fig. 3 and Table 6).

6.3. Voltage deviation Fig. 3 illustrates the work of EGWO–DA and the current model for voltage deviation. Nearly 2.003 MW is reduced by the power loss with help of the proposed model. Comparison with proper

328

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

Table 5 Comparison between algorithm in voltage deviation, two-fold objective function and fuel cost. Methods

Fuel cost ($/h)

Voltage deviation (p.u.)

Twofold objective purpose function

EGWO–DA DA-APSO [2] ABC [19] APSO [2]

790.56 802.63 803.78 803.60

0.1024 0.1164 0.1243 0.0891

808.9786 812.4526 822.3456 825.2893

Fig. 6. Voltage deviation of DA–EGWO and existing methods.

Fig. 4. Minimization of Fuel cost of Existing and Proposed methods.

Fig. 7. Convergence plot for EGWO–DA agents. Table 6 Comparison minimization power loss.

Fig. 5. Minimization of Power loss of Existing and Proposed design.

wind power system and satisfying the active power in the proposed model is taken into account in this proposed design. 6.4. Fuel cost emission The all emission natures and a suitable pricing on every kind (e.g., SOx, NOx and Cox, thermal emission, etc.) could be pioneered as an emission process. In the recent work, two important categories of emission gases are considered which are the generated NOx and SOx level. The good evaluation of the intended EGWO–DA in order to reduce the emission cost process over the 30 free trails and voltage is shown in Fig. 4.

Methods

Minimize power loss (MW)

DA–EGWO AGWO [1] PSO [8]

35 50 145

The good evaluation of the intended EGWO–DA in order to reduce the emission cost process over the 30 free trails and power loss is shown in Fig. 5 (see Figs. 6–8). Case 4: Scheduled power vs cost (wind Generation 1) This Fig. 9 illustrates the variation of the wind power cost vs. scheduled power for wind generator WG1, thus proves to be the best among all the existing methods in total cost. Case 5: Scheduled power vs cost (wind generation 2) This Fig. 10 illustrates the variation of the wind power cost vs. scheduled power for wind generator WG2, thus proves to be the best among all the existing methods in total cost. Case 6: Scheduled power vs cost (solar) This Fig. 11 illustrates the variation of the wind power cost vs. scheduled power for wind generator WG2, thus proves to be the best among all the existing methods in total cost.

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

329

Fig. 8. Convergence plot for EGWO–DA agents with Power loss.

Fig. 11. Variation of solar power cost vs schedule power for solar PV unit.

Fig. 9. Variation of wind power cost vs scheduled power for wind generator WG1.

Fig. 12. Variation of wind power cost vs Weibull scale parameter (c) for windfarm1 (bus 5).

Case 7: Probability density function parameter vs cost (wind farm # 1) This Fig. 12 illustrates the variation of the wind power cost vs. Weibull scale parameter (c) for windfarm#1 (bus 5), thus proves to be the best among all the existing methods in total cost. Case 8: Probability density function parameter vs cost (wind farm # 2) This Fig. 13 illustrates the variation of the wind power cost vs Weibull scale parameter (c) for windfarm#2 (bus 11), thus proves Fig. 10. Variation of wind power cost vs scheduled power for wind generator WG2.

to be the best among all the existing methods in total cost. Windfarm#2 (bus 11).

330

Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 319–330

Fig. 13. Variation of wind power cost vs Weibull scale parameter (c) for Windfarm#2 (bus 11).

7. Conclusion This paper presents a new hybrid EGWO–DA optimization method has been successfully implemented to resolve the OPF issue. The presented method is investigated and experimented on IEEE 30 bus system. The hybrid EGWO–DA algorithm result proves that this method is more efficient in cost reduction and power loss minimization. The presented method is evaluated with other optimization algorithms from the literature survey. The renewable energy of wind and solar of generator cost modelling is presented by comparing the costs allied with overestimation and underestimation of wind and solar power. The assessment proves that this method is more effective than the other algorithm. The presented hybrid method provides various benefits from EGWO as well as DA methods which suitable to resolve major OPF issues. Acknowledgements The authors thankfully acknowledge support from the administration, Kalasalingam Academy of Research And Education, Krishnankoil, India. The authors would like to thank the reviewers for their valuable time to review the paper and better enhancement in further. References [1] H. Xu, X. Liu, J. Su, An improved grey wolf optimizer algorithm integrated with Cuckoo search, in: Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, IDAACS, 2017 9th IEEE International Conference on, vol. 1, IEEE, 2017, September, pp. 490–493. [2] C. Shilaja, K. Ravi, Optimal power flow using hybrid DA-APSO algorithm in renewable energy resources, Energy Procedia 117 (2017) 1085–1092. [3] L. Shengsong, H. Zhijian, W. Min, A hybrid algorithm for optimal power flow using the chaos optimization and the linear interior point algorithm, in: Power System Technology, 2002. Proceedings. PowerCon 2002. International Conference on, vol. 2, IEEE, 2002, pp. 793–797. [4] A. Gacem, D. Benattous, Hybrid genetic algorithm and particle swarm for optimal power flow with non-smooth fuel cost functions, Int. J. Syst. Assur. Eng. Manage. 8 (1) (2017) 146–153. [5] D.L. Le, D.L. Ho, N.D. Vo, Hybrid differential evolution and harmony search for optimal power flow, Glob. J. Technol. Optim. 6 (2) (2015) 1–6. [6] R. Effatnejad, H. Aliyari, M. Savaghebi, Solving multi-objective optimal power flow using modified GA and PSO based on hybrid algorithm, J. Oper. Autom. Power Eng. 5 (1) (2017) 51–60.

[7] M.R. Narimani, R. Azizipanah-Abarghooee, B. Zoghdar-MoghadamShahrekohne, K. Gholami, A novel approach to multi-objective optimal power flow by a new hybrid optimization algorithm considering generator constraints and multi-fuel type, Energy 49 (2013) 119–136. [8] S. Que, D. Wu, A hybrid algorithm based on BFA and PSO for optimal reactive power problem, in: Control and Decision Conference, CCDC, 2016 Chinese, IEEE, 2016, May, pp. 1190–1193. [9] S.S. Reddy, J.A. Momoh, Minimum emissions optimal power flow in wind-thermal power system using opposition based bacterial dynamics algorithm, in: Power and Energy Society General Meeting, PESGM, 2016, IEEE, 2016, July, pp. 1–5. [10] Y. Li, Y. Li, G. Li, A two-stage multi-objective optimal power flow algorithm for hybrid AC/DC grids with VSC-HVDC, in: Power & Energy Society General Meeting, 2017 IEEE, IEEE, 2017, July, pp. 1–5. [11] M. Ghasemi, S. Ghavidel, M.M. Ghanbarian, M. Gharibzadeh, A.A. Vahed, Multi-objective optimal power flow considering the cost, emission, voltage deviation and power losses using multi-objective modified imperialist competitive algorithm, Energy 78 (2014) 276–289. [12] M. Basu, Group search optimization for solution of different optimal power flow problems, Electr. Power Compon. Syst. 44 (6) (2016) 606–615. [13] M. Güçyetmez, E. Çam, A new hybrid algorithm with genetic-teaching learning optimization (G-TLBO) technique for optimizing of power flow in wind-thermal power systems. [14] J. Radosavljević, D. Klimenta, M. Jevtić, N. Arsić, Optimal power flow using a hybrid optimization algorithm of particle swarm optimization and gravitational search algorithm, Electr. Power Compon. Syst. 43 (17) (2015) 1958–1970. [15] M. Ghasemi, S. Ghavidel, M.M. Ghanbarian, M. Gitizadeh, Multi-objective optimal electric power planning in the power system using Gaussian bare-bones imperialist competitive algorithm, Inform. Sci. 294 (2015) 286–304. [16] S.M. Mazhari, H. Monsef, R. Romero, A hybrid heuristic and evolutionary algorithm for distribution substation planning, IEEE Syst. J. 9 (4) (2015) 1396–1408. [17] A.A.A. Mohamed, A.A. El-Gaafary, Y.S. Mohamed, A.M. Hemeida, Multiobjective modified grey wolf optimizer for optimal power flow, in: Power Systems Conference, MEPCON, 2016 Eighteenth International Middle East, 2016, December, pp. 982–990. [18] H.D. Abatari, M.S.S. Abad, H. Seifi, Application of bat optimization algorithm in optimal power flow, in: Electrical Engineering, ICEE, 2016 24th Iranian Conference on, IEEE, 2016, May, pp. 793–798. [19] S. Duman, Symbiotic organisms search algorithm for optimal power flow problem based on valve-point effect and prohibited zones, Neural Comput. Appl. 28 (11) (2017) 3571–3585. [20] A. Angelis-Dimakis, M. Biberacher, J. Dominguez, G. Fiorese, S. Gadocha, E. Gnansounou, G. Guariso, A. Kartalidis, L. Panichelli, I. Pinedo, M. Robba, Methods and tools to evaluate the availability of renewable energy sources, Renew. Sustain. Energy Rev. 15 (2) (2011) 1182–1200. [21] K. Rajalashmi, S.U. Prabha, Hybrid swarm algorithm for multi objective optimal power flow problem, Circuits Syst. 7 (11) (2016) 3589.

Dr. C. Shilaja, completed her M.E. in Sathyabama university Chennai & Ph.D. in Power System at VIT University Vellore. Now currently doing POST DOCT FELLOW in Kalasalingam Academy of Research And Education Krishnankoil. She has published 12 International journals and 3 international conference. The area of interest focused on Optimal Power Flow, Power System Optimization, Renewable Energy.

Dr. T. Arunprasath, has completed his Ph.D. at Kalasalingam Academy of Research And Education Krishnankoil Tamilnadu. Currently Working in Associate professor, Department of EEE, Kalasalingam Academy of Research And Education . His area of interest focused on Medical Image processing, Image Reconstruction, PET Imaging.