1 Cost Models

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

1648 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO.

4, JULY 2015

Optimal Power Dispatch of Multi-Microgrids at


Future Smart Distribution Grids
Nima Nikmehr and Sajad Najafi Ravadanegh

AbstractIn this paper, future distribution network opera- utilization of the existing power networks, reduction of fossil-
tion is discussed under assumption of multimicrogrids (MMGs) based fuels consumption, and reduction of network losses
concept. The economic operation of MMGs is formulated as an and improve system reliability. However, higher penetration
optimization problem. A stochastically and probabilistic modeling
of both small-scale energy resources (SSERs) and load demand of SSERs causes some technical and nontechnical problems
at each microgrids (MGs) is done to determine the optimal eco- such as power quality, reliability, power management, total net-
nomic operation of each MGs with minimum cost based on the work efficiency, and interconnection of network [1]. The name
power transaction between the MGs and main grids. The bal- of the future power grid is smart grid that satisfies the electri-
ance between the total power generation in each MGs and the cal infrastructure and intelligent information networks [2], [3].
load demand is determined regarding the sold or purchase power
either by MG or by main grid. Based on the results, the mean, A MG is a section of the grid with a combination of mul-
standard deviation, and probability density function of each gen- tiply SSERs [including renewable energy sources like solar
erated power with SSERs is determined considering optimization photovoltaic (PV), wind turbine (WT)], energy storages, and
constraints. A statistical analysis for generated power and costs loads. The MG is operated in two grid-connected and isolated
is given. The power interchange between MGs is considered. The types [4]. In grid-connected mode, the MG remains connected
particle swarm optimization is applied to minimize the cost func-
tion as an optimization algorithm. Results show that it is possible to the main grid either totally or partially, and imports or
to regulate the power demand and power transaction between exports power from or to the main grid. In case of any distur-
each MGs and the main grid. Moreover, it is indicated that the bance in the main grid, the MG switches over to stand-alone
power sharing between MGs with main grid can reduce the total mode while still feeding power to the priority loads.
operation cost of the future distribution network. From the MG energy management point of view, the eco-
Index TermsEconomic dispatch, load demand, nomic scheduling of generation units, storage systems, and
microgrid (MG), particle swarm optimization (PSO), probability, loads is a crucial task, where the optimization algorithms
small-scale energy resource (SSER). can be the most important issue that can be regarded as
a major component of distribution management system, this
I. I NTRODUCTION
management is carried out using the MG central controller
UTURE distribution grids may include some micro-
F grids (MGs), which can regarded as small-scale energy
zones (SSEZs) with many uncertainties both in generation and
by receiving/sending signals to local controllers. In order to
leave behind centralized control center that makes too compli-
cated to process with variety of data and controls in MGs, [5]
load. This SSEZ can be regarded as MG. One of the basic uses decentralized multiagent platform for optimal dispatch
components of MGs is small-scale energy resources (SSERs). problem. The application of multiagent system on microgrids
Optimal operation and planning of future smart distribution operation is addressed in numerous works that makes the
grids is a challenging problem with many uncertainties. Based most benefits of system operations [6], [7]. Economic dis-
on market operation of MMGs, the optimal scheduling of patch problem can be divided into two groups. One is static
MMGs is an important topic that regulates the transaction dispatch that solve for any separated time without regard-
of the power between each MGs and main grid. The global ing the relations between time periods in order to obtaining
electrical energy demand is seen to be increasing in recent the optimal solution [8]. The another method is dynamic dis-
years. Conventional power networks are facing the prob- patch that considers a discrete time system with finite time
lems of reduction of fossil fuel resources. Wind and solar period [9], [10].
are two of the main energy sources that utilized in many There is important difference between the optimization
parts of the world. Integration of SSERs will help to better of a MG and conventional economic dispatch problem [11].
At present, there are some researches on MG optimal energy
Manuscript received April 28, 2014; revised August 4, 2014 and
November 15, 2014; accepted December 30, 2014. Date of publi- dispatch. In the grid-connected mode, the optimal energy dis-
cation February 13, 2015; date of current version June 18, 2015. patch is analyzed in MG [12] and the effect of time-of-use
Paper no. TSG-00352-2014. electricity price and electric energy transaction are considered.
The authors are with the Smart Distribution Grid Research Lab,
Department of Electrical Engineering, Azarbaijan Shahid Madani To consider the decentralized optimal power dispatch strate-
University, Tabriz 13597-45778, Iran (e-mail: [email protected]; gies in [13], the sharing of marginal cost of each power source
[email protected]). is done through iterative and communication. In [14], the real-
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. time energy optimization scheduling method is proposed in
Digital Object Identifier 10.1109/TSG.2015.2396992 independent operation mode of MG.
1949-3053 c 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
NIKMEHR AND NAJAFI RAVADANEGH: OPTIMAL POWER DISPATCH OF MMGs AT FUTURE SMART DISTRIBUTION GRIDS 1649

The economic dispatch optimization problem is solved are minimization of battery life loss, power generation costs,
with different methods in literature [15], [16]. For exam- and maximization of life time characteristics of batteries. The
ple [15] uses particle swarm optimization (PSO) algorithm multiobjective function in this paper is solved with NSGA-II
in grid-connected mode of MG and multiobjective optimiza- algorithm. Hemmati et al. [37] have minimized the costs of
tion problem without considering sold and purchase power is power generation, operation, and maintenance of MG using
regarded in economic dispatch problem [16]. Chaotic quan- a new multicross learning-based chaotic differential evolution
tum genetic algorithm [17], [18], ant colony [19], differential algorithm. This algorithm searches all dimensions of solution
evolution algorithm [20], Tabu search, genetic algorithm [21], space of continuous decision variables. The load consump-
and niching evolutionary algorithm [22] are methods that used tion and generated power by WT and PV have probabilistic
in optimal dispatching problem. behaviors in 24 h. In [38], these uncertainties are modeled
One of the important issues in the MGs is management with two point estimate method (2 PEM). Mentioned method
of load demand with power sharing capability [23]. In order uses two deterministic points on both sides of the average
to achieve a minimum operation cost and application of value () of the related distribution function for all random
SSERs with best manner, the performance of the smart energy variables. The optimization of operation costs and emission
management system is studied in [24]. Balancing of demand- is considered as multiobjective function in [39]. In this paper,
supply in MGs by using power electronic devices as controlled charging and discharging scheduling model of electric vehi-
generators with locally monitoring of frequency and voltage is cles on cost function is studied that reduce both of operation
studied in [25] and [26]. In [27], a tuning fuzzy system (TFS) cost and pollutant emissions.
is applied to improve the energy demand forecasting in MG It should be mentioned that the duty of each MG is to cre-
by a simplified economic dispatch model. The TFS is applied ate balance between supply and demand. In other word, in
to prediction model to note the actual load changes. In some simple economic dispatch problem power generation sources
paper, stochastic fluctuation of load and power regulation such as SSERs and diesel generators produce power in order
margin is applied for multiobjective economic dispatch (ED) to supply load demand and with increasing or decreasing fuel
model [28]. In [29], MG management system is extended in in diesel generators, the balance between supply and demand
a stochastic framework and uncertainties in load demand and is provided. But considering MGs concept in smart grids, the
generation by renewable sources are considered and two stage balance between generation and load is done through power
stochastic programming methods is used to optimize MG oper- exchanging between MGs as well as main grid so that the total
ations effectively. Because of uncertainties in load demand cost of power generation in each MG as well as the total cost
and intermittent behavior of power generation by PVs and of power exchanging between MGs and main grid be opti-
winds it can cause operational challenges to maintain the mized. This paper solves the optimal power dispatch problem
generation-load balance. In [30], the economic dispatch algo- considering uncertainties in load and probabilistic modeling of
rithm is solved by an unconventional, stochastic optimization generated power by renewable SSERs. In order to deal with
method. In [31], load demand management of interconnected these uncertainties, PDF is considered for power generation
MGs is described as a power dispatch optimization problem. parameters. The problem is solved with PSO algorithm. The
The demands and supplies are assumed as random variables power generation at each MG and main grid, the purchased
varying over the time but without any consideration on their and sold power by each MG and the power transaction between
probability density functions (PDFs). In [32], multiagent sys- each MG and main grid are analyzed based on operation and
tem is applied to regard a decentralized operation of MGs. maintenance (O&M) cost.
By this system, the market operation of MG model and the
interaction among MGs is modeled. Lyapunov optimization
method has answered both of MG electricity scheduling and II. MG A RCHITECTURE
energy management problems in [33]. In this paper, concept A MG is a new model for MV or LV distribution networks
of quality of service in electricity (QoSE) is the base of smart that includes SSERs and new technologies such as PHEVs and
energy management. This concept with energy storage sys- energy storages. MG networks can be connected to the main or
tems management issues are converted into stability problem external grid and supplies or absorbs power to the load demand
using both of QoSE and battery virtual queues. Based on MG both in MG or main grid. In this paper, a distribution network
concept and also because of penetration of SSER into power with three MGs is analyzed. Each MG or SSEZ consists of
networks, the unit commitment (UC) problem is encountered three SSERs. The SSER units which is used include PV, WT,
with new challenges. Zhao et al. [34], using duality-based micro gas turbine (MT), fuel cell (FC), and combined heat and
method, first transformed UC problem into a convex opti- power (CHP). The structure of SSEZs is indicated in Fig. 1.
mization problem and then subgradient-based algorithm is
proposed to achieve the optimal UC solution. In [35], the
proposed MG-based co-optimization planning problem and III. P ROBABILISTIC A NALYSIS M ETHOD
annual reliability sub-problem consider as mixed-integer pro- At present, high penetration of SSERs into distribution grids
gramming problem. Minimization of total system planning affect operation and planning of the power systems. In WTs
cost is assumed as cost function. The proposed model for and PV power generation, wind speed, and solar radiation
random component outages is analyzed with numerical sim- are prime energy sources, respectively. Because of stochastic
ulations method such as Monte Carlo method. Goals of [36] behavior of wind speed and sun irradiance, power generation
1650 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 4, JULY 2015

by (2). In this circumstance, load of each MG has random


variable behavior.
2) SSERs Power Generation Model:
a) WT power generation model: Generated power by
WT depends on wind speed. The wind speed varies every
minute, hour, day, and season of the year, which highlights
the importance of a probability model. The Weibull dis-
tribution is used to represent the distribution for the wind
speed for long-term planning purposes [42]. Weibull PDF is
as follows:
  1

e( ) v 0
v
fv (v) = v
(2)
Fig. 1. Structure of network with three MGs (SSEZ). 0 otherwise.
In (2) , , and v are shape parameter of Weibull function,
of the above energy resources undergoing significant uncer- scale parameter of this function, and wind speed, respectively.
tainties. Uncertainties analysis of the impact of SSERs such The wind speed is prime energy source and must be converted
as WT and PV units on current power systems based on into power based on
deterministic methods is complicated scenario. Although the  n 
v vn
deterministic analysis of power systems is a common used in Pr vcutout
n v n vcutin v vr
cutin r
electric utilities, the probabilistic analysis of power systems PG (v) = Pr vr v vcutout (3)


at presence of uncertainties is a very powerful tool for opti- 0 otherwise
mal operation and planning of power systems. In probabilistic
analysis, input data have PDF and these data can be described where PG (v) and Pr are generated power of WT at speed v,
by cumulative distribution function (CDF). Consequently, the respectively. vcutin , vcutout , and vr are WT parameters.
obtained results from probabilistic analysis are also presented Theses parameters are low cut, high cut, and rated speed of
in PDF and CDF forms. Probabilistic analysis is applied WT. Parameter n describes the rate of characteristic curve
in calculation of probabilistic load flow (PLF) [40] in that between vcutin and vr . Maximum value of generated power
authors are performed a good review on PLF problem. It by WTs is considered 250 kW.
is also applied into short-term and long-term planning [41]. b) PV power generation model: Solar radiation and air
One of the problems which has stochastic and probability temperature are two important parameters for power genera-
behavior at presence of SSERs is ED. Considering uncer- tion in PV. These parameters are variable in any time of day.
tainties in load consumption and generated power by gen- In this paper, irradiance and air temperature are modeled by
eration units, ED problem is analyzed using probabilistic normal distribution function for a given time. The properties
method. of the PV in operation conditions differ from the standard con-
In this paper, PDF is considered for input data such as load, dition (GSTC = 1000 W/m2 , Tc = 25 C). The PV modules are
generated power by SSERs, purchased and sold powers by tested at standard test condition (STC). The output power of
MGs, cost of transaction, and costs O&M. Hence, the output the module can be calculated as follows [43]:
results are represented in framework of PDF and CDF. GING
Ppv = PSTC (1 + k(Tc Tr )) (4)
GSTC
IV. S YSTEM M ODEL where Ppv and PSTC are output power of the module at irradi-
A. Power Model ance GING and rated power at STC respectively. Tc and Tr are
1) Load Demand Model: In deterministic load model for cell and air temperature, respectively. k is maximum power
a MG, it is supposed that the load of each MG is constant temperature coefficient. The rated power of PVs is 250 kW.
at any study period. Because of variable nature of load, it is
suitable to model the stochastic behavior of the load. This B. Cost Model
modeling can be achieved either by measured data analysis or 1) WT and PV Cost Model: The cost of primary energy
by application of mathematical model. The MG load behavior often determines the cost of generated power by SSERs. In
is modeled as a normal distribution function. In this case, the WT and PV, wind and solar radiation are primary energies
MG load is defined by a mean value () and standard deviation respectively with zero fuel costs. For the cost function of WT
( ) as and PV units, only O&M costs are considered. In this paper,
 
1 (Pl )2 the costs of O&M are supposed 0.1095 ($/kWh) and 0.1368 for
f (Pl ) = exp . (1) PVs and WTs, respectively at every samples. It should be men-
2 2 2
tioned that all samples have been considered only for one-time
To model the load variation in a given period, a predefined interval. The costs of O&M (CO&M ) for WT and PV are as
number of load samples based on normal distribution with follows:
mean and standard deviation are generated to model the
load uncertainty. The MGs loads at each sample are calculated CO&M, WT, s = KO&M Ps . (5)
NIKMEHR AND NAJAFI RAVADANEGH: OPTIMAL POWER DISPATCH OF MMGs AT FUTURE SMART DISTRIBUTION GRIDS 1651

TABLE I
C HARACTERISTICS OF G ENERATION U NITS

Ps is generated power of PV or WT units considering are able to inject power into grid, which in that, the MGs take
sample s. money from network. By selling power to grid, the costs of
2) FC Cost Model: In FC, hydrogen is combined with oxy- MGs operation are reduced that is an economical operation.
gen and electricity, heat, and water is produced. Produced If one of the MGs be unable to supplies own load demand, it
electricity is a dc current and to connect this unit into MG, it can purchase electrical power from main grid. The purchased
is necessary to use power electronic devices. FC can produce and sold cost of MGs is described as
electricity as long as fuel is being provided. The efficiency of
FC (s ) in sample s depends on output power in same sam- Costpur,sm = c Pbuy,smi (11)
i
ple. The efficiency of FC will be lower if the output power be
increased. The cost of generated power by FC achieve from Costsell,sm = d Psell,smi . (12)
Cnl Ps i
CostFC,s = . (6)
L s In the above equations, Costpur,sm and Costsell,sm are pur-
In this paper, Cnl and L are 0.76 $/m3 and 9.7 kWh/m3 . chased and sold powers costs of MG m in sample s, respec-
These parameters describe natural gas price value ($/m3 ) and tively. Pbuy,smi and Psell,smi are purchased and sold powers by
low-hot value (kWh/m3 ), respectively. The O&M costs for FC ith unit of MG m in sample s, respectively. Based on Fig. 1,
are calculated based on (6). The value of KO&M for FC is each MG consist three units. The cost of transaction of powers
considered 0.016 ($/kWh). between MG and external grid is described as
3) MT Cost Model: MTs are small high-speed gas turbines. Costtrans,ms = Costpur,ms Costsell,ms . (13)
The size of MT is from 25 to 500 kW [45]. A MT supplies
input mechanical energy for the MT generator system, which In the proposed paper, c and d are purchased price and sold
is converted by the generated to electricity energy. Unlike FC, price, respectively. The value of c and d, respectively, are 0.16
the efficiency of MT increases with increase of supplies power. and 0.12 ($/kWh) for all MGs. Table I shows some mentioned
The cost function of generated power of MT is calculated characteristics of MGs.
by (7) and O&M costs are obtained by (6) for each sample.
The value of KO&M is considered 0.088 ($/kWh) for MT V. P ROBLEM F ORMULATION
Cnl Ps The MG optimization problem is rather similar to the ED
CostMT,s = . (7)
L s with traditional generation units. The purpose of ED prob-
4) CHP Cost Model: CHP system is an economical type lem is minimization of cost function of generation considering
of power delivery method that provides electricity and heat equality and inequality constraints.
at the same time. In MT units with CHP performance, the
efficiency of MT increase and the fuel cost of MT decrease A. Objective Function
to a great extent. The fuel cost of MT with CHP performance
ED is a nonlinear problem. Objective function includes gen-
is as follows:
erated power, purchased and sold power, and O&M costs. In
CostCHP,s = CostMT,s BCHP,s (8) this problem, the cost of power generation must be minimized.
rec (T,s e,s ) The total energy cost for providing energy to consumers is
BCHP,s = CostMT,s (9)
b equal to the summation of the energy costs of MGs. The
where BCHP,s is cost reduction of generated power by MT in objective function is as follows:
sample s, because of using exhaust gas heat. rec is heat recov-
Min : OF = Fs (14)
ery factor. T , e , and b are total efficiency of CHP, electrical s
efficiency of MT and efficiency of boiler, respectively. On the
other hand, composition of (8) and (9) concludes as Fs = Costgen,si + (c Pbuy,sm d Psell,sm )
  i m
rec (T,s e,s ) + CostO&M,s (15)
CostCHP,s = CostMT,s 1 . (10)
b Costgen,si = Costgen,MT,si
CostMT,s obtains from (7). In this paper, rec and b are
+ Costgen,FC,si + Costgen,CHP,si (16)
assumed to be 0.95 and 0.80, respectively.
5) Modeling of Purchased and Sold Powers Cost: In this CostO&M,s = CostO&M,WT,s + CostO&M,PV,s + CostO&M,MT,s
paper, all MGs are connected into main grid. So, all of them + CostO&M,FC,s + CostO&M,CHP,s . (17)
1652 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 4, JULY 2015

Fig. 2. Path of the particles movement.

In (14), OF is objective function that is summation of func-


tion F. Function F is optimized for each samples separately
by PSO algorithm. Indeed F is a [1 s] function or matrix. In
this paper, Costgen,si is the cost of generated power buy unit i
at sample s.

B. Problem Constraints
1) Power Balance Constraint: The total real-power gener-
ation plus purchased power from external grid must balance
the predicted power demand plus the sold power to main grid,
at any sample of PDF

Pl,s + Psell,sm = Pgen,si + Pbuy,sm . (18)
m i m

2) Inequality Constraint: In this paper, upper and lower


constraint is considered for generated power of SSERs. These
constraints are applied into power transaction between MGs Fig. 3. Flowchart of implementation of PSO on probabilistic ED.
and main grid
TABLE II
Min,gen < Pgen < PMax,gen
Psi si si
(19) PSO PARAMETERS U SED FOR S YSTEMS
Min,pur < Ppur < PMax,pur
Psm sm sm
(20)
Psm
Min,sell < Psm
sell < Max,sell .
Psm (21)
The optimal dispatch problem in (14) is minimized with
PSO algorithm considering equality and inequality constraints. where
c1 , c2 acceleration constant in the range [0, 2];
VI. R EVIEW OF PSO A LGORITHM r1 , r2 uniform random value in the range of [0, 1];
The main idea of the PSO algorithm is inspired first by Pbesti best previous position of particle 3i;
Kennedy and Eberhart [46] in 1995. The idea is suggested gbest best particle among all Pbest ;
from offensive particles movement such as birds or fish. When Si , xi velocity and position vectors of particle i,
the birds want to move, they use present position and neigh- respectively;
bors birds position in order to reach bird with the best k number of iterations;
position. In PSO algorithm, instead of birds or fish, particles do W inertia weight factor;
this paper. According to Fig. 2, any of particles are described Iter current iteration number;
with two vectors Si and xi . In every movement steps of parti- W min ,W max minimum and maximum inertia weights factor.
cles population, every particle is updated by two values. First PSO algorithm can be described as follows.
value is the best previous position of particle i that called p- Step 1 (Initialization): The velocity and position vectors are
best and evaluated by using fitness function. Second value is initialized randomly.
the best particle among all p-bests that called gbest . By finding Step 2 (Update Velocity): The velocity vector of the all
these values, any of particles update their new velocity and particles is updated by using equation.
position by following relations: Step 3 (Position Update): The position vector of the all
particles is updated by using equation.
Sik+1 = W Sik + c1 r1 pbesti xi Step 4 (Memory Update): Pbesti and gbest are updated as
+ c2 r2 (gbest xi ) (22) follows:
xik+1
= xik + Sik+1 (23)
(Wmin Wmax ) If : F ( pi ) < F pbesti ; pi pbesti (25)
W = Wmax iter (24) If : F ( pi ) < F (gbest ); pi gbest . (26)
itermax
NIKMEHR AND NAJAFI RAVADANEGH: OPTIMAL POWER DISPATCH OF MMGs AT FUTURE SMART DISTRIBUTION GRIDS 1653

TABLE III
S TATISTICAL A NALYSIS OF G ENERATED P OWERS AND C OSTS

Fig. 4. Load demand of MGs 1, 2, and 3 and network.

Fig. 5. Probabilistic modeling of PV and WT primary input and generated


The flowchart of solving probabilistic optimal power dis- power in MG1.
patch problem with PSO is illustrated in Fig. 3 graphically.
According to the flowchart, a PDF function is used for load, TABLE IV
wind speed, solar irradiation, and air temperature, so these VALUES OF PARAMETERS IN WT AND PV
input variables are defined as probabilistic variables. Based
on correlation between network input and output variables, the
output of the network parameters show probabilistic behavior.
To explain this, for example, Fig. 5 is illustrated the proba-
bilistic modeling of both wind speed and solar irradiation and
consequently the probabilistic modeling for wind and solar based on measured data of 12 h. This problem is solved for
power in MG1. only 12 h of the day. Based on the above discussion, there
are approximately 300 (3 93 + 3 7 = 300) samples for
VII. S IMULATION R ESULTS AND D ISCUSSION each hours.
In this section, distribution of power problem among loads The load demand of each MGs and main grid is illustrated
of MGs is tested on proposed MMG network. Fig. 1 shows in Fig. 4.
the structure of MMG network with three MGs. Each MG In this paper, each MG is modeled as a small disco with
consists three SSERs to power generation. All basic data for capability of power interchange between each MG and also
simulation are mentioned in the rest of this paper. between each MG and the main grid. To consider the effect
The aim of this paper is to solve the optimal power dispatch of power generation cost in each MG, it is assumed that the
of MMGs at the presence of SSERs such as WT, PV, FC, and characteristic of SSERs such as size, and technology of WT,
CHP for a given hours. Because of the intermittent behavior PV, and FC is different. Based on the above discussion, the
of some renewable energy resources and load variation this cost of per kilowatt generated power by each SSERs at each
paper is done based on uncertainty in input data. MG is different. Indeed, this paper is modeled as a multi-
To model the uncertainty of wind speed and solar irradiation area economic dispatch problem in market operation of power
the historical measured data for 100 days (93 days of summer system that each MG is seem as a virtual power plant. In this
and the first week of autumn) with three samples for each case, the probabilistic economic dispatch problem is solved
hours (3 93 + 3 7 samples) at 12 h of each day, is used using PSO as a heuristic optimization algorithm. The charac-
as practical data to model the distribution function of wind teristics of PSO algorithm is described in Table II. Based on
speed and solar irradiation. As an economic dispatch problem this figure and Table III, the mean cost of generated power in
this paper is solved the MMG economic dispatch problem MG1 is lower than MG2 and MG3, because MG1 has two DG
1654 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 4, JULY 2015

TABLE V
S TATISTICAL A NALYSIS OF P OWER T RANSACTION AND C OSTS

TABLE VI
S TATISTICAL A NALYSIS OF P OWER T RANSACTION AND C OSTS IN I NTERCONNECTED MMG S M ODE

In this paper, ED problem is solved considering PDF


for some parameters of power generation. Proposed example
solves ED problem at presence of load demand and renew-
able SSERs uncertainties. The results includes the total cost
at any states of generated and transaction powers. The prob-
abilistic analysis about ED has a good prospect in operation,
planning, unit commitment problem studies, etc. PDF of the
purchased and sold powers of MGs gives an insight vision
to the dispatcher to evaluate the risk of change in system
total cost with respect the variation in load and SSER power.
This value at risk for purchase and sold power can be cal-
culated and can be useful for dispatchers. The mean value
of the PDF can use as the power with high probability of
the appearance. Indeed, in this case, it is possible to evalu-
ate the forecasting errors in load, wind, and solar short-term
scheduling.
Fig. 6. PDF of generated power of MGs.
In Table IV, the values of parameters of WT and PV units
are given that are used in simulation.
units, WT and PV, and these units have no cost about power Table V describes the purchased and sold powers, costs
generation. Besides, the sum of generated power by MGs in of purchased and sold power, and transaction costs between
Fig. 6 must be equal with total power generation of network MGs and external grids based on mean and standard devi-
at any sample. ation of samples. In this table to calculate the transac-
Table III describes the powers, costs of power, and O&M tion cost, (14) is used for each MG. The purchased and
costs of each units based on mean and standard deviation of sold powers of MGs are indicated in Fig. 7. The CDF
300 samples. of total costs of network as PDF and CDF are shown
The density of generated power at each MGs, is shown in in Fig. 8.
Fig. 6. To consider the power transfer capability at intercon-
Using obtained costs of generated power, O&M, and trans- nected MG, the simulation is repeated and the results are
action powers between main grid and each MG, the total cost given in Figs. 9 and 10. Moreover, in Table VI, the detail
of each MG can be described as follows for any sample: of the analysis is given. For this case, the power transac-
tion between two MGs as well as each MG and the main
CostMG = Costgen,i + (c Pbuy,m d Psell,m ) grid are given. Comparing Tables V and VI is clear that
i m total cost of operation is lower in case of interconnected
+ CostO&M . (27) MMG operation.
NIKMEHR AND NAJAFI RAVADANEGH: OPTIMAL POWER DISPATCH OF MMGs AT FUTURE SMART DISTRIBUTION GRIDS 1655

Fig. 7. PDF of purchased and sold powers.

TABLE VII
S TATISTICAL A NALYSIS OF P OWER T RANSACTION AND C OSTS IN MMG S M ODE W ITH L OSSES

Fig. 8. PDF and CDF of total cost of network.

Fig. 10. PDF of sold power in different states of MG1 in interconnected


MMG mode.

size of the network. To evaluate the effects of network losses


in the proposed method, in Table VII the statistical analysis
of power transaction and costs in multimicrogrids (MMGs)
mode considering network loss is given. In this table, the net-
work main parameters are compared. Based on Table VII, the
cost of power generation within each MG is increased because
of the loss modeling, while the cost of purchased power is
decreased. The PDF of the network loss is shown in Fig. 11.
Fig. 9. PDF of purchased power in different states of MG1 in interconnected According to the figure while the input parameters have the
MMG mode. random behavior, hence the network output parameters such
as network loss is also a random parameter. The histogram of
Because of the small size of the load and power gener- the network loss is indicated in this figure. This figure show
ation the effect of network loss can be considered in this that the network loss can be modeled as a normal distribution
paper. The network loss is also small because of the small function.
1656 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 4, JULY 2015

[6] P. Jain and S. J. Ranade, Capacity discovery in customer-driven micro-


grids, in Proc. North Amer. Power Symp. (NAPS), Starkville, MS, USA,
Oct. 2009, pp. 16.
[7] Y. Xu and W. Liu, Stable multi-agent-based load shedding algorithm for
power systems, IEEE Trans. Power Syst., vol. 26, no. 4, pp. 20062014,
Nov. 2011.
[8] M. A. Abido, Multi objective evolutionary algorithm for electric
power dispatch problem, IEEE Trans. Evol. Comput., vol. 10, no. 3,
pp. 315329, Jun. 2006.
[9] M. Basu, Dynamic economic emission dispatch using nondominated
sorting genetic algorithm-II, Int. J. Elect. Power Energy Syst., vol. 30,
no. 2, pp. 140149, 2008.
[10] L. Xiaoping, D. Ming, H. Jianghong, H. Pingping, and P. Yali, Dynamic
economic dispatch for microgrids including battery energy storage, in
Proc. 2nd IEEE Int. Symp. Power Elect. Distrib. Gen. Syst. (PEDG),
Hefei, China, Jun. 2010, pp. 914917.
[11] C. A. Hernandez-Aramburo, T. C. Green, and N. Mugniot, Fuel con-
sumption minimization of a microgrid, IEEE Trans. Ind. Appl., vol. 41,
Fig. 11. PDF of network loss. no. 3, pp. 673681, May/Jun. 2005.
[12] N. Ming, H. Wei, G. Jiahuan, and S. Ling, Research on economic
operation of grid-connected microgrid, Power Syst. Technol., vol. 34,
no. 11, pp. 3842, 2010.
VIII. C ONCLUSION [13] K. De Brabandere, K. Vanthournout, J. Driesen, G. Deconinck, and
R. Belmans, Control of microgrids, in Proc. IEEE Power Eng. Soc.
In this paper, future distribution networks operation is dis- Gen. Meeting, Tampa, FL, USA, Jun. 2007, pp. 17.
cussed in the presence of MMGs is discussed. The economic [14] S. Qingjun, G. Guangchao, and J. Quanyuan, Real-time energy dispatch
operation of MMGs is formulated as an optimization problem. of standalone microgrid, in Proc. Chin. Soc. Elect. Eng. (CSEE), 2012,
pp. 2635.
A stochastically and probabilistic modeling of both SSERs and [15] M. Yuyang, L. Jinling, and Z. Guodong, Improved multi-objective
load is performed to determine the optimal operation of each particle optimization algorithm based scheduling optimization of grid-
MGs with minimum cost based on economic analysis of the connected microgrid, Elect. Power Sci. Eng., vol. 28, no. 7, pp. 1520,
2012.
power transaction between the MGs and main grid. The bal- [16] F. A. Mohamed and H. N. Koivo, Online management of microgrid with
ance between the total power generation in each MGs and battery storage using multiobjective optimization, in Proc. 1st Int. Conf.
the load demand is determined regarding the sold or purchase Power Eng. Energy Elect. Drives (POWERENG07), Setubal, Portugal,
Apr. 2007, pp. 231236.
power of the main grid. Based on the results, the mean, stan- [17] G. Liao, Solve environmental economic dispatch of smart micro-
dard deviation, and PDF of each generated power with SSERs grid containing distributed generation systemUsing chaotic quantum
is determined considering optimization constraints. Statistical genetic algorithm, Int. J. Elect. Power Energy Syst., vol. 43, no. 1,
pp. 779787, Dec. 2012.
analysis for generated power and costs are given. The PSO is [18] G. Liao, Using chaotic quantum genetic algorithm solving environ-
applied to minimize the cost function as an optimization algo- mental economic dispatch of smart microgrid containing distributed
rithm. Results show that it is possible to regulate the power generation system problems, in Proc. Int. Conf. Power Syst. Technol.,
Hangzhou, China, Oct. 2010, pp. 17.
demand and transaction between each MG and the neighbors [19] A. Xin, C. Minyong, and L. Zhili, Based on chaos ant colony algo-
MG and between each MG and the main grid. Moreover, it is rithm for the micro-grid environmental and economy operation, J. Univ.
indicated that the power sharing between MGs with main grid North China Elect. Power J., vol. 36, no. 5, pp. 26, 2009.
[20] M. Varadarajan and K. S. Swarup, Solving multi-objective optimal
can reduce the total operation cost of the future distribution power flow using differential evolution, IET Gen. Transmiss. Distrib.,
network. One of the main results of this paper by probabilis- vol. 2, no. 5, pp. 720730, Sep. 2008.
tic modeling of the input variables, the output variables can [21] Y. Kinjyo et al., Optimal operation of smart grid with fuel cell in
isolated islands, J. Int. Council Elect. Eng., vol. 2, no. 4, pp. 423429,
be represented as random variables. This leads to a better 2012.
and comprehensive vision for network experts to manage the [22] S. Conti, R. Nicolosi, S. A. Rizzo, and H. H. Zeineldin, Optimal dis-
marginal operation of the network under uncertainties. This patching of distributed generators and storage systems for MV islanded
microgrids, IEEE Trans. Power Del., vol. 27, no. 3, pp. 12431251,
can guarantee the robust operation of smart distribution grids Jul. 2012.
in the presence of network uncertainties. [23] I. Koutsopoulos and L. Tassiulas, Challenges in demand load control for
the smart grid, IEEE Netw., vol. 25, no. 5, pp. 1621, Sep./Oct. 2011.
[24] C. Chen, S. Duan, T. Cai, B. Liu, and G. Hu, Smart energy management
R EFERENCES system for optimal micro grid economic operation, IET Renew. Power
Gener., vol. 5, no. 3, pp. 258267, 2011.
[1] N. Saito, T. Niimura, K. Koyanagi, and R. Yokoyama, Trade-off analy- [25] J. Guerrero, M. Chandorkar, T. Lee, and P. Loh, Advanced control archi-
sis of autonomous microgrid sizing with PV, diesel, and battery storage, tectures for intelligent microgridsPart I: Decentralized and hierarchical
in Proc. IEEE Power Energy Soc. Gen. Meeting, Calgary, AB, Canada, control, IEEE Trans. Ind. Electron., vol. 60, no. 4, pp. 12541262,
Jul. 2009, pp. 16. Apr. 2013.
[2] A. Ipakchi and F. Albuyeh, Grid of the future, IEEE Power Energy [26] J. M. Guerrero, P. C. Loh, T.-L. Lee, and M. Chandorkar, Advanced
Mag., vol. 7, no. 2, pp. 5262, Mar. 2009. control architectures for intelligent microgridsPart II: Power quality,
[3] The U.S. Department of Energy. (Oct 2014). The Smart Grid: energy storage, and AC/DC microgrids, IEEE Trans. Ind. Electron.,
An Introduction. [Online]. Available: http://www.oe.energy.gov/ vol. 60, no. 4, pp. 12631270, Apr. 2013.
Documents and Media/DOE_SG_Book_Single_Pages(1).pdf [27] T. S. Mahmoud, D. Habibi, O. Bass, and S. Lachowicz, Tuning
[4] H. Jiayi, J. Chuanwen, and X. Rong, A review on distributed energy fuzzy systems to achieve economic dispatch for microgrids, in Proc.
resources and microgrid, Renew. Sustain. Energy Rev., vol. 12, no. 9, IEEE Conf Innov. Smart Grid Technol. (ISGT), Perth, WA, Australia,
pp. 24722483, Dec. 2008. Nov. 2011, pp. 16.
[5] N. Cai, N. T. T. Nga, and J. Mitra, Economic dispatch in microgrids [28] Y. Wei, Optimal scheduling for distributed energy resource,
using multi-agent system, in Proc. North Amer. Power Symp. (NAPS), M.A. thesis, Dept. Modern Phys., Hefei Univ. Technol., Hefei, China,
Champaign, IL, USA, Sep. 2012, pp. 15. 2010.
NIKMEHR AND NAJAFI RAVADANEGH: OPTIMAL POWER DISPATCH OF MMGs AT FUTURE SMART DISTRIBUTION GRIDS 1657

[29] A. Parisio and L. Glielmo, Stochastic model predictive control [43] T. Ishii, K. Otani, T. Takashima, and S. Kawai, Estimation of the max-
for economic/environmental operation management of microgrids, imum power temperature coefficients of PV modules at different time
in Proc. Eur. Control Conf. (ECC), Zrich, Switzerland, Jul. 2013, scales, Solar Energy Mater. Solar Cells, vol. 95, no. 1, pp. 386389,
pp. 20142019. Jan. 2011.
[30] R. H. Liang and J. H. Liao, A fuzzy-optimization approach for gen- [44] N. Augustine, S. Suresh, P. Moghe, and K. Sheikh, Economic dispatch
eration scheduling with wind and solar energy systems, IEEE Trans. for a microgrid considering renewable energy cost functions, in Proc.
Power Syst., vol. 22, no. 4, pp. 16651674, Nov. 2007. IEEE PES Innov. Smart Grid Technol. (ISGT), Washington, DC, USA,
[31] M. Fathi and H. Bevrani, Statistical cooperative power dispatching in Jan. 2012, pp. 17.
interconnected microgrids, IEEE Trans. Sustain. Energy, vol. 4, no. 3, [45] W. G. Scott, Micro-turbine generators for distribution systems, IEEE
pp. 586593, Jul. 2013. Ind. Appl. Mag., vol. 4, no. 4, pp. 5762, May/Jun. 1998.
[32] M. A. Lpez, S. Martn, J. A. Aguado, and S. de la Torre, Market- [46] J. Kennedy and R. Eberhart, Particle swarm optimization, in Proc.
oriented operation in microgrids using multi-agent systems, in Proc. Int. Conf. Neural Netw., Perth, WA, Australia, Nov./Dec. 1995,
Int. Conf. Power Eng., Energy Elect. Drives (POWERENG), Mlaga, pp. 19421948.
Spain, May 2011, pp. 16.
[33] Y. Huang, S. Mao, and R. M. Nelms, Adaptive electricity scheduling
in microgrids, IEEE Trans. Smart Grid, vol. 5, no. 1, pp. 270281,
Jan. 2014.
[34] B. Zhao, Y. Shi, X. Dong, W. Luan, and J. Bornemann, Short-term Nima Nikmehr was born in Iran, in 1990. He
operation scheduling in renewable-powered microgrids: A duality based received the B.Sc. degree in electrical engineering
approach, IEEE Trans. Sustain. Energy, vol. 5, no. 1, pp. 209217, from Azarbaijan Shahid Madani University, Tabriz,
Jan. 2014. Iran, in 2012. He is currently pursuing the M.Sc.
[35] A. Khodaei and M. Shahidehpour, Microgrid-based co-optimization of degree in electric power engineering from Islamic
generation and transmission planning in power systems, IEEE Trans. Azad University, Tehran, Iran.
Power Syst., vol. 28, no. 2, pp. 15821590, May 2013. His current research interests include evolution-
[36] B. Zhao, X. Zhang, J. Chen, C. Wang, and L. Guo, Operation opti- ary algorithms and intelligence computing in power
mization of standalone microgrids considering lifetime characteristics systems, probabilistic modeling in power systems,
of battery energy storage system, IEEE Trans. Sustain. Energy, vol. 4, operation, and economic aspect of microgrids.
no. 4, pp. 934943, Oct. 2013.
[37] M. Hemmati, N. Amjady, and M. Ehsan, System modeling and
optimization for islanded micro-grid using multi-cross learning-based
chaotic differential evolution algorithm, Int. J. Elect. Power Energy
Syst., vol. 56, pp. 349360, Mar. 2014.
[38] A. Baziar and A. Kavousi-Fard, Considering uncertainty in the opti-
mal energy management of renewable micro-grids including storage
devices, Renew. Energy, vol. 59, pp. 157166, Nov. 2013. Sajad Najafi Ravadanegh was born in Iran, in
[39] A. Zakariazadeh, S. Jadid, and P. Siano, Multi-objective scheduling of 1976. He received the B.Sc. degree from the
electric vehicles in smart distribution system, Energy Convers. Manage., University of Tabriz, Tabriz, Iran, and the M.Sc.
vol. 79, pp. 4353, Mar. 2014. and Ph.D. degrees from the Amirkabir University of
[40] P. Chen, Z. Chen, and B. Bak-Jensen, Probabilistic load flow: Technology, Tehran, Iran, in 2001, 2003, and 2009,
A review, in Proc. 3rd Int. Conf. Elect. Utility Dereg. Restruct. Power respectively, all in electrical engineering.
Technol. (DRPT), Nanjing, China, Apr. 2008, pp. 15861591. He is currently an Assistant Professor with
[41] M. Leite da Silva, S. M. P. Ribeiro, V. L. Arienti, R. N. Allan, and the Electrical Engineering Department, Azarbaijan
M. B. Do Coutto Filho, Probabilistic load flow techniques applied to Shahid Madani University, Tabriz, where he is
power system expansion planning, IEEE Trans. Power Syst., vol. 5, responsible for the Smart Distribution Grid Research
no. 4, pp. 10471053, Nov. 1990. Laboratory. His current research interests include
[42] Z. Bie, G. Li, H. Liu, and X. Wang, Studies on voltage fluctuation smart distribution networks and microgrids optimal operation and planning,
in the integration of wind power plants using probabilistic load flow, power system stability and control, power system controlled islanding, opti-
in Proc. IEEE Power Energy Soc. Gen. Meeting Convers. Del. Elect. mization algorithms applications in power systems, nonlinear dynamic, and
Energy 21st Century, Pittsburgh, PA, USA, Jul. 2008, pp. 17. chaos. He has authored/co-authored of over 40 technical papers.

You might also like