Optimizing Combined Emission Economic Dispatch For Solar Integrated Power Systems
Optimizing Combined Emission Economic Dispatch For Solar Integrated Power Systems
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
Abstract—The dispatch of power at minimum operational smooth convex cost functions and efficient solutions were
cost of thermal energy sources has been a significant part proposed [6]. However, the complex nature of the commonly
of research since decades. Recently with increasing interests used generation facilities has made the cost functions essen-
in renewable energy resources, the optimal economic dispatch
has become a challenging issue. This paper presents combined tially non-convex [7]. For instance, thermal units equipped
emission economic dispatch (CEED) model for a solar photo with multi-valve steam turbines produce ripples in the cost
voltaic (PV) integrated power system with multiple solar and functions. These ripples like effects are originated due to
thermal generating plants. We formulate mixed integer binary opening or closing of valves at various stages, and are known
programming problem (MIBP) subject to various practical con- as the valve point effects (VPEs). In addition to VPEs, the
straints. A decomposition framework is proposed where the
original problem is split into two sub-problems. Particle swarm aspect of multiple fuel options also results in non-smooth ob-
optimization (PSO), Newton-Raphson method, and binary integer jective functions for ED optimization [8]. In literature, various
programming (BIP) techniques are exploited to find the joint artificial intelligence (AI) based optimization techniques were
optimization solution. The proposed model is tested on IEEE 30 proposed to address the complex ED problem e.g., genetic
bus system. The simulation results demonstrate the effectiveness algorithm (GA) [9], particle swarm optimization (PSO) [10],
of the proposed model.
Index Terms—Economic Dispatch, Power Systems, Photo neural network (NN) [11], evolutionary programming (EP)
Voltaic, PSO [12] and tabu search (TS) [13].
PSO has been the most popular technique to address ED
I. I NTRODUCTION optimization due to its simplicity and ability to solve the
In recent decades, deployment of renewable energy sources complex problems [14]. In [15], the authors addressed non-
in place of thermal generation has received a lot of attention. convex problem by a hybrid scheme which combines PSO and
This shift of momentum took place due to the rapidly depleting sequential quadratic programming (SQP). The constraints like
reserves of fossil fuels, the dramatic increase in fuel prices, and power balance, security, power generation limits, ramp rate
the environmental concerns associated with thermal fuels [1]. limits, and spinning reserve were involved. PSO was used as
Economic dispatch (ED) is a well known optimization problem a main optimization tool and SQP as a local optimizer for
that aims to find an appropriate combination of power shares further tuning of the obtained solution. In [16], constriction
from committed generating units that results in minimum factor PSO (CFPSO) had been proposed for non-convex ED
cost. There are numerous constraints involved in ED problem under various constraints such as generator limits and power
such as power limits of generators, power balance, prohibited balance. The inclusion of constriction factor ensured fast con-
operating zones, ramp rate limits etc. Various optimization vergence of the algorithm by gradually reducing the velocities.
techniques have been reported in literature to solve the ED Considering more complex cost function under the prohibited
problem [2]. operating zones and VPEs, a technique named ‘iteration PSO
ED with thermal generating units has been one of the most with time varying acceleration coefficients (IPSO-TVAC)’ was
frequently investigated problems in power system optimization proposed in [17]. The iteration term in the velocity equation
[3]– [4]. The structure of fuel cost functions for thermal enhanced the memorizing ability of PSO whereas the time
generating units plays a significant role in ED problem [5]. varying acceleration coefficients resulted in better balance
Traditionally, the problem had been extensively studied with between social and local components.
Thermal generating units produce emissions that result in
This work was supported in part by the National Natural Science Foundation serious environmental impacts [18]. Therefore, along with the
of China under Grant 61422109 amd Higher Education Commission (HEC) cost minimization, a significant intention has been paid to
Pakistan.
N. A. Khan is with the Department of Electrical Engineering, COMSATS keep the emissions at minimum level. A multi-objective ED
Institute of Information Technology, Park Road, Chak Shahzad, Islamabad, problem, which involves both the fuel cost and emissions
Pakistan (Email: [email protected]). is known as combined emission economic dispatch (CEED).
G. A. S. Sidhu (Corresponding Author) is with the Department of Electrical
Engineering, COMSATS Institute of Information Technology, Park Road, Selvakumar et al. [19] solved the CEED problem under the
Chak Shahzad, Islamabad, Pakistan (Email: [email protected]). constraints of power balance and power generation capacity
F. Gao is with Department of Automation, Tsinghua University, State Key limits using PSO. The authors of work in [20] proposed
Lab of Intelligent Technologies and Systems, Tsinghua National Laboratory
for Information Science and Technology (TNList) Beijing, P. R. China, (email: multi-objective particle swarm optimization (MOPSO) and
[email protected]). used a diversity-preserving mechanism to find the wide range
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
of pareto-optimal solutions. Moreover, [20] considered the generator limits, renewable energy limits, and the ramp rate
practical security constraints which were missing in [19]. In limits. However, the practical security and the network losses
[21], the authors enhanced MOPSO procedure by redefining constraints were missing and the authors did not test the pro-
the local and global best individuals resulting in significant posed solution on any real network. More recently, the work in
improvements. In [22], multi-objective CEED problem was [32] developed a rule-based power management algorithm to
converted into a single objective optimization by introducing achieve dispatch characteristics similar to conventional thermal
a price penalty factor. The problem was solved using the generating units. However, the study was limited to only a
artificial bee colony (ABC) algorithm. Further, a modified single PV plant without implementation on real network. Thus,
harmony search algorithm to solve the combined economic various constrains of practical power system were absent.
and emission load dispatch (CEELD) problem was designed In this paper, we present a joint CEED optimization problem
in [23]. which aims to 1) minimize the thermal fuel cost, the emissions,
and the cost of solar generation, 2) maximize the share of
The ED of renewable integrated power system becomes the solar power and the number of participating solar units1 .
more complex due to intermittent and uncontrollable nature We consider various practical network and security constraints
of the renewable energy. The work in [24] proposed a two- i.e. power balance, generator bounds, bus voltage bounds,
step framework to evaluate the energy storage in a wind aided reactive power limits, line thermal capacity limits, available
power system. The authors dealt with the wind intermittency solar power, and upper bound on solar penetration. It should
using reserve generation. However, this work was limited to be noted that the existing works in literature may only have
battery storage scenario and the actual network constraints some of these constraints. The contributions of this paper are
were missing. In [25], the authors studied short term planning summarized as:
of virtual power plants (VPP) under the objectives of max- • A joint multi-objective integer optimization problem
imizing economic profit and minimizing deviations between is formulated to minimize the overall operational
actual and contracted energy. The scope of this paper was cost/emissions and to maximize the total share/number
little to propose bid based operation of VPPs connected to a of solar power plants.
very small network. The authors in [26] presented a stochastic • A decomposition framework is first adopted to split the
model to solve an optimal power flow (OPF) problem for original problem into two sub-problems. Then, a mixed
wind integrated power systems. In this paper, the level of optimization scheme is proposed which exploits the PSO,
wind penetration was determined based on the demand-side the Newton-Raphson method, and the binary integer
flexibility. The work [24]– [26] did not consider the emissions programming.
and important constraints like voltage limits, tap limits, reac- • The proposed model is employed to IEEE 30-bus network
tive power limits etc., were also ignored. In [27], the authors to verify its practical application.
provided economic environmental dispatch (EED) model by • Through simulations, we show the better performance of
integrating the renewable energy sources in conventional grid. the proposed algorithm in terms of cost minimization and
The test system involved six thermal machines, one photo maximization of the solar share.
voltaic (PV) plant, and a single wind turbine. The problem
The rest of the paper is organized as follows: system
was formulated subject to power balance, generator limits,
model and joint optimization problem are presented in section
and renewable energy limit constraints. A strength pareto
II. Section III elaborates the proposed solution. Simulation
evolutionary algorithm (SPEA) was used as an optimization
results and discussions are given in section IV, and finally the
tool. This work was extended in [28] and a dynamic economic
conclusion is presented in section V.
emission dispatch (DEED) model was proposed considering
the additional constraints of security, ramp rate limits, and line II. S YSTEM M ODEL AND P ROBLEM F ORMULATION
capacity. However, the considered test system was very small, A. System Model
with a limited penetration of renewable energy share and had
We consider a power generation supply system consisting
only six buses. Moreover, the constraints such as reactive
of n number of thermal and m number of solar generating
power limits, bus voltage limits, and tap changer limits were
plants2 , supplying N B number of loads as shown in Fig. 1.
neglected. The work [29] proposed a hybrid model for off-grid
Let Pi be the power generated by the ith thermal generating
applications to minimize the fuel and the battery wear costs
unit, the fuel cost in $/h is expressed as [8]:
subject to various constraints like availability of the PV power,
the battery bank charge, and the load power demand. In [30], Fi (Pi ) = ai Pi2 + bi Pi + ci + |ei sin(fi × (Pimin − Pi ))|,
the authors investigated ED with high penetration of renewable (1)
energy resources and evaluated the effects of uncertainties and
1 Maximizing the number of participating solar plants minimizes the prob-
network congestion on the dispatch strategies. Further, [31]
considered the solar integrated power system and proposed an ability of failure due to intermittent nature of solar radiation.
2 We assume that a feasible solution always exists in terms of demand and
efficient PSO based solution to a more general CEED opti- supply, that is to say the generating plants are assumed to satisfy the demands
mization problem subject to the constraints of power balance, of users in the entire system.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
NB
X
Pk = |Vk ||Vs ||Yk,s |cos(θk,s − ζk + ζs ), (8)
s=1
NB
X
Qk = |Vk ||Vs ||Yk,s |sin(θk,s − ζk + ζs ), (9)
s=1
Fig. 1: System Model. where P k and Qk are the real and reactive power injections at
th
bus k, respectively, Yk,s is the (k, s) element of admittance
matrix Y , θk,s is the angle of the admittance element Yk,s ,
where ai , bi , ci , ei , and fi are fuel cost coefficients of ith ζ. is the angle of respective voltage, and |.| represents the
generating unit. Pimin is the minimum power limit of ith magnitude. Tap changing transformers for voltage adjustments
generating unit. Thermal generating units produce emissions are equipped with limited number of taps. Therefore the
which can be presented in kg/h by: selected tap Tt of tth transformer should be within the range
of available taps:
Ei (Pi ) = αi Pi2 + βi Pi + γi + εi exp(δi × Pi ), (2)
where αi , βi , γi , εi , and δi are emission coefficients of the Ttmin ≤ Tt ≤ Ttmax ∀ t = 1, 2, ...., N T, (10)
ith generating unit. The sine and exponential terms in (1) and
(2) represent VPEs. with N T be the total number of transformers. Also the power
Let P gsj be the power generated by the j th solar plant, the flow Slz (k, s) at any line z connected between buses k and s
cost of operation in $/h is given by following expression, should not exceed its capacity, i.e.,
Gj (P gsj ) = P U Costj × P gsj , (3)
Slz (k, s) ≤ Slzmax (k, s) ∀ z = 1, 2, ...., N Z, (11)
th
where P U Costj is the per unit cost in $/M W h of j solar
plant. P gsj is given by [27]:
where N Z is total number of lines. Slz (k, s) can be obtained
Si as:
P gsj = P ratedj{1 + (Tamb − Tref )× α pvj }× , (4)
1000
with P ratedj being the rated power of j th solar plant, Slz (k, s) =Vk × (Vk − Vs )∗ × (Yk,s )∗
Tamb the ambient temperature, Tref the reference temperature, + Yk,0 × Vk2 , (12)
α pvj the temperature coefficient, and Si the incident solar
radiation.
Each thermal generating unit has its own range of real power where (x)∗ represents the complex conjugate of x.
generation Pi and reactive power generation Qi , i.e.,
Qmin ≤ Qi ≤ Qmax ∀i = 1, 2, ..., n, (6) This work deals with the CEED problem incorporating both
i i
the thermal and solar power generations. It aims to reduce the
where Pimax , Qmin
i , and Qmax
i are the maximum real power overall costs and the thermal emissions. Another target is to
limit, minimum reactive power limit, and maximum reactive get the maximum benefits from the installed solar capacity by
power limit of the ith generating unit, respectively. maximizing the solar share and to decrease the probability of
All the generating plants and loads are assumed to be con- solar failure by increasing the number of participating solar
nected to IEEE 30-bus network, which has its own limitations. units. The multi-objective optimization can be formulated,
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
with
where c1 is a vector of cost coefficients, I1 is identity vector,
xvy+1 =ω y vtsyx + C1 r1 (Pbestyx − Ptsyx )+ and c2 is a vector of solar powers. This is a standard inte-
C2 r2 (gbesty − Ptsyx ), (37) ger programming problem and Binary Integer Programming
toolbox in MATLAB is used to obtain the solution. The
where xvy+1 is velocity of xth particle at iteration y +1, toolbox takes the coefficients of objective function as well
ω y is a parameter known as inertia weight at iteration y, as of constraints as input and provides the binary decision
C1 and C2 are acceleration coefficients, whereas r1 and variables and the corresponding fitness evaluation as output.
r2 are two random numbers between 0 and 1. This completes the solution of optimization in (13). For
(5) Update the position of each particle as more clarity, the detailed steps of proposed joint optimization
solution are summarized in algorithm 1.
Ptsy+1
x
max(xxy+1 , Ptsmin
x ), if xxy+1 < Ptsmax
x , Algorithm 1 Proposed Joint Optimization Solution
=
min(xxy+1 , Ptsmax
x ), if xx y+1
> Ptsmax
x , 1: Initialize swarm size, ωmin , ωmax , ω y = ωmax , itermax ,
(38) vxmin , and vxmax . Generate P̂i , Pˆsk , and vˆi according to
the equations (26), (27), and (28). Set y = 1.
where xxy+1 is given by, 2: Generate Pyswarm and vswarmy
, such that the equations
xx y+1
= Ptsyx + vtsy+1 . (39) (5), (19), and (28) are satisfied.
x
3: Obtain the initial power flows from Newton Raphson
(6) Find the power flows and evaluate the fitness method as discussed in III-A.
Fyswarm = F1 (Pts) and update Fgbest = 4: After step 8 of the solution proposed in III-A, set Pi∗ = P̂i
minx (Fyswarm ). Obtain gbest and pbest as and P s∗k = Pˆsk .
5: For the obtained P s∗k , solve the sub-problem II in (46)
using Binary Integer Programming toolbox of MATLAB.
gbest =Pyswarm (indg ), (40) 6: Terminate the procedure and display the results.
Pxbest =PTBx (indxp ), (41)
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
9
The proposed model is implemented on standard IEEE 30 10:00 Hrs.
11:00 Hrs.
bus system with 6 thermal machines and 13 solar plants. The 8 12:00 Hrs.
data for the thermal units and the power demand, for the 13:00 Hrs.
IEEE 30 bus system, and for the solar plants are taken from 7 14:00 Hrs.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
350 1.12
10:00 Hours
11:00 Hours 1.1
300 12:00 Hours
250
15:00 Hours 1.06
200 1.04
1.02
150
1
100
0.98
50 0.96
0.94
0 0 50 100 150 200
Iteration
−50
0 5 10 15 20 25 30 35 40 Fig. 4: Voltage magnitude vs iterations for the demand at time
Line Number
12:00 hours.
Fig. 3: Capacities and power flows of the lines.
V. C ONCLUSION
This work presented a new CEED model for a solar PV
integrated power system which could minimize the fuel cost,
emissions, solar cost, and could maximize the solar share
and the number of solar plants. The constraints of power
Figure 5 shows the total solar cost, the solar share, and the balance and the bounds on generators, renewable energy,
number of selected plants in different hours. It can be noticed voltage magnitudes, transformer taps, and line capacities were
that the total solar cost may come out to be different for the considered in the joint optimization problem. The amount
same number of solar plants and is directly proportional to of solar share was subject to various network constraints,
the solar share. For instance, the number of solar plants at renewable energy limit, and available solar power. A hybrid
11:00 hours and 12:00 hours are same; however, the cost at optimization scheme was proposed using the PSO, the Newton
the 12:00 hours is less as compared to 11:00 hours. This is method, and the binary integer programming. The results
because of the selection of different solar plants (i.e., 3, 4, 5, showed the promising performance of the proposed algorithm
6, 7, 8, and 9 instead of 4, 5, 6, 8, 9, 10, and 13 ) resulting satisfying all constraints with an acceptable complexity in
in different total share. Lastly, we analyze the generation of terms of convergence.
emissions with the variations in solar and the thermal shares in
Fig. 6. As expected, the emissions increase with the increase in R EFERENCES
thermal share and vice verse. Similarly, the emissions decrease
[1] B. Bose, “Global energy scenario and impact of power electronics in
with an increase in solar share, i.e., the amount of emission is 21st century,” IEEE Trans. Ind. Electron., vol. 60, no. 7, pp. 2638–2651,
inversely proportional to the total solar share. 2013.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
[6] S. Yang, S. Tan, and J.-X. Xu, “Consensus based approach for economic
dispatch problem in a smart grid,” IEEE Trans. Power Syst., vol. 28,
300 no. 4, pp. 4416–4426, 2013.
Solar Share
(MW)
10
optimization for various types of economic dispatch problems,” Int. J.
5 Electr. Power Energy Syst., vol. 28, no. 1, pp. 36–42, 2006.
[11] K. Y. Lee, A. Sode-Yome, and J. H. Park, “Adaptive hopfield neural
0 networks for economic load dispatch,” IEEE Trans. Power Syst., vol. 13,
10 11 12 13 14 15 no. 2, pp. 519–526, 1998.
4
x 10
[12] Y. M. Park, J. R. Won, and J. B. Park, “A new approach to economic load
10 dispatch based on improved evolutionary programming,” Eng. Intell.
Syst. Elect. Eng. Commun., vol. 6, no. 2, pp. 103–110, 1998.
Solar Cost
economic dispatch problems,” Int. J. Electr. Power Energy Syst., vol. 42,
500 no. 1, pp. 508–516, 2012.
[18] M. Brown and S. Ulgiati, “Emergy evaluations and environmental
0 loading of electricity production systems,” J. Clean. Prod., vol. 10, no. 4,
10 11 12 13 14 15
pp. 321–334, 2002.
300
[19] A. I. S. Kumar, K. Dhanushkodi, J. J. Kumar, and C. K. C. Paul, “Par-
Solar share
1000
tal/economic dispatch problem,” Electr. Power Syst. Res., vol. 79, no. 7,
500 pp. 1105–1113, 2009.
[22] Y. Sonmez, “Multi-objective environmental/economic dispatch solution
0
10 11 12 13 14 15 with penalty factor using artificial bee colony algorithm,” Scientific
Time of the day (h) Research and Essays, vol. 6, no. 13, pp. 2824–2831, 2011.
[23] B. Jeddi and V. Vahidinasab, “A modified harmony search method for
Fig. 6: Level of emissions for various thermal and solar shares. environmental/economic load dispatch of real-world power systems,”
Energy Convers. Manag., vol. 78, pp. 661–675, 2014.
[24] N. Li, C. Uckun, E. M. Constantinescu, J. R. Birge, K. W. Hedman, and
A. Botterud, “Flexible operation of batteries in power system scheduling
[2] A. J. Wood and B. F. Wollenberg, Power generation, operation, and with renewable energy,” IEEE Trans. Sustain. Energy, vol. 7, no. 2, pp.
control. John Wiley & Sons, 2012. 685–696, 2016.
[3] I. Farhat and M. El-Hawary, “Interior point methods application in [25] F. Luo, Z. Y. Dong, K. Meng, J. Qiu, J. Yang, and K. P. Wong, “Short-
optimum operational scheduling of electric power systems,” IET Gener. term operational planning framework for virtual power plants with high
Transm. Distrb., vol. 3, no. 11, pp. 1020–1029, 2009. renewable penetrations,” IET Renewable Power Generation, vol. 10,
[4] S. Kar, G. Hug, J. Mohammadi, and J. M. Moura, “Distributed state no. 5, pp. 623–633, 2016.
estimation and energy management in smart grids: A consensus innova- [26] W. A. Bukhsh, C. Zhang, and P. Pinson, “An integrated multiperiod opf
tions approach,” IEEE J. Sel. Topics Signal Process., vol. 8, no. 6, pp. model with demand response and renewable generation uncertainty,”
1022–1038, 2014. IEEE Trans. Smart Grid, vol. 7, no. 3, pp. 1495–1503, 2016.
[5] W. T. Elsayed and E. F. El-Saadany, “A fully decentralized approach [27] S. Brini, H. H. Abdallah, and A. Ouali, “Economic dispatch for power
for solving the economic dispatch problem,” IEEE Trans. Power Syst., system included wind and solar thermal energy,” Leonardo J. Sci.,
vol. 30, no. 4, pp. 2179–2189, 2014. vol. 14, pp. 204–220, 2009.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2587665, IEEE Access
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.