Kumar 2019

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

2019 2nd International Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT)

Solution Approach to Unit Commitment


Problem Using GAMS Environment
Vineet Kumar1, R. Naresh2, Amita Singh3
National Institute of Technology Hamirpur
1,2,3

knocking at local optima and may fail to provide best


solution. Secondly, all classical approaches are based on an
Abstract—In day to day life, with an ever-growing demand
in power sector, planning and operation plays a vital role in
providing an economical, reliable and efficient electricity to the assumption that the objective function to be handled is
consumers. In this regard, unit commitment (UC) plays a
continuous and differentiable in nature whereas a practical
significant part in daily planning and optimal scheduling of
power system network is more complex in nature.
generating units so as to meet the hourly load demand in an
efficient manner. This paper focusses on presenting a robust
Contemporary evolutionary techniques possess an advantage
and effective methodology for solving the UC problem using of being versatile in handling qualitative constraints but their
GAMS simulation environment. In this work, to assess the main drawback lies in the fact that their computational time
effectiveness of GAMS over MATLAB environment, 3 and 4 increases exponentially as the problem dimension increases
thermal generating units with and without spinning reserves and also their time to convergence is uncertain (convergences
and ramp rate constraints have been considered over 24-hour is guaranteed) (20-23). Thus, a robust and effective approach
time horizon. is needed to attain an optimal as well as realistic results in
lesser computational time and burden while satisfying all the
Keywords— Unit commitment, Priority list (Pl), General diverse constraints. These constraints bring in complexity and
Algebraic Modeling System (GAMS) difficulty while solving solution of UC problems. Thus, the
practical UC problem has high dimensionality and non-
I. INTRODUCTION linearity, and hence requires optimal, robust and fast solution
methodology that can handle all the challenges present in the
Over the past few years, there has been an expeditious UC problem.
rise in energy demand thus eventually leading to a great deal
of pressure on power utilities to generate more power at This paper presents the GAMS software as an
lowest possible cost. The generating units need to be optimization approach for solving UC problem. The work
optimally scheduled to meet the load demand while fulfilling focusses on solving the UC problem while considering time
both unit and system constraints. This scheduling process is constraints i.e. up and down time, load balance constraint,
widely known as unit commitment problem. The objective of system spinning reserve, generation limits, ramp rate limits,
the unit commitment is to determine the on/off status of initial status of units etc.
generating units and corresponding least cost generation
schedule satisfying load demand over the planning horizon. II. UNIT COMMITMENT PROBLEM FORMULATION
A variety of versions of unit commitment are used to
minimize the cost like UC integration with renewable energy A number of operating units has been taken into
resources, profit based unit commitment etc. Power balance consideration for proper scheduling and operation of UC
is also important in UC that satisfy the demand supply of problem in power system. The foremost goal of UC is to
customers. The main objective of UC problem is to minimize evaluate the optimal on/off operation of fuel units
the total generation cost (production cost, shutdown and start- throughout the definite time period in a manner that the total
up costs) while satisfying the various unit and system generating cost (TGC) being a combination of production
constraints (1). The main constraints include time constraints, cost, start up and shut down cost is lowest while fulfilling
load balance constraint, unit generation limits, system the diverse system and unit constraints (3).
spinning reserve, ramp rate limits, initial status etc.
Combinative nature and high extensity of the UC problem The cost function conforming to the production cost can
restricts the efforts to proceed with any detailed and precise be approximated to have a quadratic cost characteristic.
mathematical approach that are capable of unravelling the Mathematically, objective function is represented by eq. (1)
entire problem for any real-sized power system problem. as (12):
Many scholars have done immense research and have applied N T
TGC= min   {Fi (Pit ) Uit +Uit (1-Uit-1 ) SUCi,t +Uit-1(1-Uit ) SDCi,t } (1)
various approaches to solve UC problem. These approaches i=1t=1
can be classified as (a) Classical approaches e.g. Priority where,
method (2), Dynamic programming (3), Lagrange relaxation
method (4), Branch and Bound method (5), mixed integer Fi (Pit )= ai +bi Pit +ci (Pit )2
linear programming (6-7) etc. (b) Evolutionary approaches where
e.g. Particle swarm optimization (8-10, 16), Ant colony
system (11-13), artificial bee colony (14) genetic algorithm
HSC , if T
i  Tt  T +T 
 i,down i,off i,down i,cold 
(15, 19), Harmony search method (17), other metaheuristic SUCi,t =  
t
 CSCi , if Ti,off >Ti,down +Ti,cold 
approaches (18) etc.. From the above literature review, it is  
concluded that main drawback of classical approaches is

978-1-7281-1772-0 ©2019 IEEE

Authorized licensed use limited to: Murdoch University. Downloaded on June 14,2020 at 00:59:32 UTC from IEEE Xplore. Restrictions apply.
where, total generation cost is denoted by TGC, SDC is shut Pi,tminU it  Pit  Pi,tmaxU it (8)
down cost and treated as constant value and SUC is start-up
cost which relies on units which are off prior to on. In this
paper, various constraints considered in UC problem are where, Pi,tmax is maximum generation limit and Pi,tmin
A. System balance power constraint is minimum generation limit
At each hour, the output power generated by committed III. METHODOLOGY
units must fulfill the power demand in the forecasted period
This paper focuses on three approaches to explain the
of time and is given as (13):
UC problem that are described below as:
N t t t
 PU -P = 0; t = 1,2,...T (2)
i=1 i i load A. Dynamic Programming Approach

B. Ramp rate constraint


The application of dynamic programming (DP) to
The maximum limit by which the generator output optimization problem do not require calculation of
power can be augmented in a certain time duration is derivatives. Therefore, many practical small sized problems
determined by ramp-based constraint and is represented as having nonlinear and nonconvex objective functions can be
(12): solved. The major disadvantage of DP is curse of
t t-1 dimensionality in general the complicated UC problem is
Pi - Pi  UR(i) when production increases (3) simplified when solved with DP by assuming following
t-1 t assumptions (1):
Pi - Pi  DR(i) when production decreaes (4)
• Define A state contains array of elements with
stated units working and other are off-line.
C. Unit minimum up and down time constraints
• Independency on time for a start-up cost.
• For shutting down of a generating unit there is no
The time constrained unit commitment (TCUC)
problem includes non-linear constraints like minimum up costs.
and down time for every unit. Thus, it indicates that when in • It follows a severe priority order, and in every time
running mode, a unit must not be shut down immediately interval there must be a definite minimum amount
and once in off state, it must not be turned on without a of capacity that is always operating.
minimum delay in time period, as expressed below(13):
On arranging a DP process, to run behind in time interval,
t-1 t-1 t initial from the last period to be considered, posterior to the
(Ti,on - Ti,up )×(Ui -Ui )  0; (5) starting hour. The recursive process to compute the minimum
t-1 t t-1 cost in hour K with grouping I is,
(Ti,off - Ti,down )×(Ui -Ui )  0; (6)
Fcost (K, I)  min{L}[Pcost (K, I )  Fcost (K 1, L)  Scost (K 1, L : K, I )] (9)
i = 1,2,..., N; t = 1,2,...T
where
Pcos t ( K , I )  production cost for state transition (K, I )
D. Spinning reserve
Scost ( K  1, L : K , I )  conversion cost from state transition (K - 1,L)
Spinning reserve is expressed as a pre-stated percentage
or a certain proportion of the predicted maximum demand, to state case (K,I )
and it should be obtainable during the scheduling and Fcost ( K , I )  minimum total cost to reach at state case (K, I )
planning period [39]. It is supplied by the devices that are
synchronized to the power network and are capable of
affecting the generator’s watt power and is expressed by the In forward DP method, we stated an approach as the
eq (7) shown below (12): transition case, or route, from one case to another time
N max t horizon at a given time period.
t t
 Pi,t Ui  Pload + SR (7)
i=1 X = States to search every period

where, is maximum generation limit, is N = Plan, or routes, to save at every step


power demand and is system spinning reserve at time
period t. Also, the spinning reserve capacity is considered B. Lagrange Relaxation method
during on state of a generating unit. When compared with dynamic programming the
solution of UC problem is made simple by Lagrangian
E. Unit generation limit constraint relaxation (LR). LR method offers a quick solution because
of the dual optimization behavior of technique. We start by
Generation limit is used to commit a unit which
defining the variable (1).
must be within specified limits as shown below [12]:

Authorized licensed use limited to: Murdoch University. Downloaded on June 14,2020 at 00:59:32 UTC from IEEE Xplore. Restrictions apply.
t th XPRESS, SNOPT, etc. The main idea behind the
U i =0 if i unit is OFF / shut down at time t evolution/development of this software were people of
t th mathematics who relied on optimization tools, as a more
U i =1 if i unit is ON / running at period t prominent and sophisticated framework for handling real
time problems in engineering and technology. UC problem
Cost function and numerous constraints of the UC is a extremely nonlinear and mixed integer problem that can
problem are stated as: be modelled using GAMS, a quite useful approach to
establish an accurate model. Simultaneous development,
Loading Constraints: solution and maintenance of models is possible by
maintaining the same GAMS model file [24]. Sets, data,
N t t variable, equation, model and output forms the basic structure
t of a mathematical model. Different solvers are present in
Plaod -  Pi U i =0
i 1 GAMS software and according to problem formulation and
Unit Limits: constraint, we used solvers and get better results in compare
to other intelligent methods [25]. The assumption made in
t min t t max the GAMS is the existence of a fixed lower and upper limit
U i Pi  Pi  U i Pi for i = 1... N, and t = 1...T on nonlinear expressions in the NLP or MINLP to be solved.
Figure 1 shows the complete layout process in GAMS
Minimum up and down-time variables can also be included software.
in the problem.

Objective function is: Define optimization problem and model design in


GAMS
T N t t t t
  [Fi (Pi )+ Start up costi,t ] U i = F(Pi ,U i ) (10)
t=1i=1
Model explanation, pre- solution report
By relaxing the coupling constraint, the Lagrange processing, appropriate solver in GAMS
relaxation method decomposes the overall UC problem into
number of single unit problems over complete time horizon.
In dual optimization the dual function of the UC problem is
first maximized with respect to Lagrange multiplier and then Nonlinear optimization, followed by local and global
primal value of the cost function is minimized with regard to search method
power output parameter of the generating units. The relative
duality gap is stated as the ratio of the change between primal
to dual value of the objective function to the dual value.
Unnecessary switching in and switching off of some units Optimal calls for other programs and solver
may give rise to sensitivity issues. Fig. 1. Solution procedure in GAMS

q ( )  max t q (  )

where , q(λ) is dual gap. IV. RESULTS AND DISSCUSIONS
q (  )  min t t  ( P , U ,  ) In this section, two case studies have been examined to
Pi Ui assess the effectiveness of the GAMS-DICOPT solver over
MATLAB. The effectiveness of the GAMS software for UC
Initially  is taken as zero, then adjusting  by equation:
is tested for two case studies. The first case study comprises
d of 3 and 4 thermal generating units. The thermal unit
t  t  [ q ( )] parameter as well as the hourly load demand for case study 1
d
where and 2 is adopted from (1) and (11), respectively. The hourly
load demand is mentioned in table 1 and 2. The thermal unit
d parameter and cost coefficients are listed in table 3 and 4.
  0.01 where q (  ) is positive
d The program file is written in GAMS 25.1 and executed on
Intel core i5, 4 GHz processor PC.
d
  0.002 where q ( ) is negative
d Table 1. Load demand schedule for 3 thermal generating units.
C. GAMS Software
Time (Hour) 1 2 3 4
t
GAMS is mostly alike to fourth generation grouped Pload (MW) 170 520 1100 330

language and it was first algebraic language. It has an


inbuilt IDE (integrated development environment) which is
linked to variety of third-party solvers easily. The solvers Table 2. Load demand schedule for 4 thermal generating units.
present in GAMS are BARON, CONOPT, COIN-OR
solvers, CPLEX, Gourbi, MOSEK, DICOPT, SULUM, Time
1st 2nd 3rd 4th 5th 6th 7th 8th
(Hr.)

Authorized licensed use limited to: Murdoch University. Downloaded on June 14,2020 at 00:59:32 UTC from IEEE Xplore. Restrictions apply.
Pload
400 530 600 540 400 280 290 500 GAMS 20161 5.1324
(MW)

Table 3. Characteristics limits for 3 thermal generating units. Table 6. Result of 4 generating units

bi Software Cost (Rs) CPU Time (sec)


Unit Pimax Pimin ai ci
(Rs/MWh)
(i) (MW) (MW) (Rs/hr) (Rs/MW2h)
DP [1] 74110 9.9648
1 600 100 500 10 0.002
2 400 100 300 8 0.0025 GAMS 73982 8.4532
3 200 50 100 6 0.005

B. Case study 2
Table 4. Characteristics limits for 4 thermal generating units
GAMS-DICOPT solver has been applied on 4 thermal
Unit Pimax Pimin Ti,up Ti,down HSCi CSCi Ti,cold generating units throughout 24-hr scheduled time period
(i) (MW) (MW) (h) (h) (Rs) (Rs) (hr) with 1-hour time horizon, considering system generation
1 80 25 4 2 150 350 4
constraints, spinning reserve, ramp rate constraints, unit
2 250 60 5 3 170 400 5
constraints, and power balance constraints. Optimum UC
3 300 75 5 4 500 1100 5 operation for 4 generating unit, 24-hr test system in the
4 60 20 1 1 0 0.02 0 presence of ramp rate limit is presented in Table 7.
A. Case study 1 Operating cost result of this case is compared with results of
other methods reported in literature and are mentioned in
GAMS-DICOPT solver has been applied on 3 and 4
Table 8. The operating cost obtained using GAMS for this
thermal generating units.
case is Rs. 26854.2 which is less as compared to DP and ant
Case study 1 (a)- The UC problem has been solved by colony system. From table 8 it can be observed that in case
applying LR approach on 3 thermal generating units using of 4 generating units, the operating cost using GAMS-
GAMS-DICOPT solver, considering generation constraints, DICOPT solver is 0.489% less than DP and 0.251% less
unit limits and minimum up and down time limits. than ACS [11]. Total time taken to execute the UC problem
Case study 1(b)- The UC problem has been solved by GAMS-DICOPT solver is more as compared to other
applying DP approach on 4 thermal generating units using methods mentioned in table 8. In UC, time does not play a
GAMS-DICOPT solver, considering all diverse system significant role because scheduling and planning is done
constraints. before 24-hour time period. So, in this paper operating cost
In order to evaluate the superiority of the presented has been considered as a major concern.
methodology, a comparative analysis is executed between
Table 7. Optimum UC operation for 4-unit, 24-hr test study in the
the results obtained in MATLAB environment (1) and presence of ramp rate limit
GAMS-DICOPT environment in terms of operating cost Time
(Rs.) and CPU time (sec). A relative performance analysis 1st Unit 2nd Unit 3rd Unit 4th Unit
(Hr.)
for 3 and 4 generating units has been presented in table 5 1 1 1 1 1
and table 6, respectively. From table 5 it can be observed 2 1 1 1 1
that in case of 3 generating units using LR approach, the 3 1 1 1 1
operating cost using GAMS-DICOPT solver is 0.0099% less 4 1 1 1 1
than that of MATLAB (1). Similarly, the execution time
5 1 1 1 1
obtained using GAMS IS 5.1324 sec which is less as
6 1 1 1 0
compare to (1).
7 1 1 1 0
From table 6 it can be observed that in case of 4
generating units using DP approach, the operating cost using 8 1 1 1 1
GAMS-DICOPT solver is 0.172% less than that of 9 1 1 1 1
MATLAB (1). Similarly, the execution time obtained using 10 1 1 1 1
GAMS is 8.4532 sec which is less as compare to (1). 11 1 1 1 1
12 1 1 1 1
Table 5. Result of 3 generating units 13 1 1 1 1
Software Cost (Rs) CPU Time (sec) 14 1 1 1 1
15 1 1 1 1
LR [1] 20163 7.4742 16 1 1 1 1
17 1 1 1 1

Authorized licensed use limited to: Murdoch University. Downloaded on June 14,2020 at 00:59:32 UTC from IEEE Xplore. Restrictions apply.
18 1 1 1 1 [8] Juste K. A., Kita H., Tanaka E. and Hasegawa J., “An evolutionary
programming solution to the unit commitment problem”, IEEE Trans.
19 1 1 1 1 Power Syst., Vol. 14, No. 4, pp. 1452–1459, 1999.
20 1 1 1 1 [9] M. N. C. Othman, T. K. A. Rahman, H. Mokhlis and M. M. Aman,
21 1 1 1 1 “Solving unit commitment problem using multi-agent evolutionary
22 1 1 1 1 programming incorporating priority list”, Arab J. Sci, Eng., Vol. 40,
pp. 3247-3261, 2015.
23 1 1 1 1
[10] Khorasani J., “A new heuristic approach for unit commitment problem
24 1 1 1 1 using particle swarm optimization”, Arab. J. Sci. Eng., Vol. 37, pp.
1033–1042, 2012.
[11] Sishaj P. Simon, Narayana Prasad Padhy, R.S. Anand, “An ant colony
Table 8. Comparison of cost obtained with other methods system approach for unit commitment problem”, Electrical Power and
Energy System, Vol. 28, pp. 315–323, 2006.
Dynamic Ant Colony [12] Prateek K. Singhal, R. Naresh and Veena Sharma, “A Novel Strategy-
Methods Programming System GAMS Based Hybrid Binary Artificial Bee Colony Algorithm for Unit
[11] [11] Commitment Problem”, Arab J Sci Eng., Vol. 40, pp. 1455-1469,
2015.
Cost (Rs.) 26986.4 26921.94 26854.2 [13] Prateek K. Singhal, R. Naresh and Veena Sharma, “A modified binary
artificial bee colony algorithm for ramp rate constrained unit
commitment problem”, Int. Trans. Electr. Energy Syst., Vol. 25, pp.
Time 3472-3491, 2015.
2.04 3.42 10.45
(Sec.)
[14] Karaboga D. and Basturk B., “A powerful and efficient algorithm for
numerical function optimization: artificial bee colony (ABC)
V. CONCLUSSION algorithm”, J. Global Optim., Vol. 39, pp. 459–471, 2007.
[15] Kazarlis S.A., Bakirtzis A.G., Petridis V.: A genetic algorithm
An efficient GAMS environment was successfully solution to the unit commitment problem. IEEE Trans. Power Syst.
applied on UC problem with diverse constraints. The Vol. 11, No. 1, pp. 83–92, 1996.
introduced Gams software outperforms the results obtained [16] Najafi A., Farshad M. and Falagi H., “A new heuristic method to
in MATLAB environment in terms of accuracy, high solve unit commitment by using time variant acceleration coefficients
efficiency and robustness. It can also be understood that for particle swarm optimization algorithm”, Turk. J Elec. Eng. & Comp.
Sci., Vol. 23, pp. 354-369, 2015.
a system under study, GAMS produces the same solution at
[17] Pourjamal Y. and Ravadanegh S. N., “HSA based solution to the UC
each run of the software whereas, after every run of the problem”, Int. J. Electr. Power Energy Syst., Vol. 46, pp. 211–220,
algorithm a metaheuristic approach results in a different 2013.
solution. The simulation results of GAMS shows a [18] D. F. Rahman, A. Viana and J. P. Pedroso, “Metaheuristic search-
prodominant approach for dealing UC problem in power based methods for unit commitment”, Int. J. Electric Power Energy
sector. Syst., Vol. 59, pp. 14-22, 2014.
[19] Hatim S. Madraswala and Anuradha S. Deshpande, “Genetic
REFERENCES Algorithm Solution to Unit Commitment Problem”, 1st IEEE
International Conference on Power Electronics. Intelligent Control
[1] G Wood, J., Wollenberg, F.: Power Generation Operation and and Energy Systems, 2016.
Control, 2nd edn. Wiley Publication, India, 2016. [20] Wenting Hou, Rujie Zhu, Hua Wei, Hiep TranHoang, “Data-driven
[2] Senjyu, T., Shimabukuro K., Uezato K. and Funabashi T., “A fast affinely adjustable distributionally robust framework for unit
technique for unit commitment problem by extended priority list”, commitment based on Wasserstein metric”, IET Generation,
IEEE Trans. Power Syst., Vol. 18, No. 2, pp. 882–888, 2003. Transmission & Distribution, Vol. 13, Issue 6, pp.no. 890-895, 2019.
[3] Ouyang Z. and Shahidehpour S. M., “An intelligent dynamic [21] Zeinab G. Hassan, M. Ezzat, Almoataz Y. Abdelaziz, “Solving unit
programming for unit commitment application”, IEEE Trans. Power commitment and economic load dispatch problems using modern
Syst., Vol. 6, No. 3, pp. 1203–1209,1991. optimization algorithms”, International Journal of Engineering,
Science and Technology, Vol.9, No.4, pp. 10-19, 2017.
[4] Feng X. and Liao Y., “A new lagrangian multiplier update approach
for Lagrangian relaxation-based unit commitment”, Electr. Power [22] Fouad R. Zaro, Salah J. Alqam, “Solving Dynamic Load Economic
Compon. Syst., Vol. 34, pp. 857–866, 2006. Dispatch Using GAMS Optimization Algorithm”, IEEE Jordan
International Joint Conference on Electrical Engineering and
[5] Chen C. L. and Wang S. C., “Branch-and-bound scheduling for
Information Technology (JEEIT), 2019.
thermal generating units”, IEEE Trans. Energy Convers., Vol. 8, No.
2, pp. 184–189, 1993. [23] Abarghooee R. A., Niknam T., Bavafa F. and Zare M., “Short term scheduling
of thermal power systems using hybrid gradient based modified teaching–
[6] Li X., Li T., Wei J., Wang G. and Yeh W.W.G., “Hydro unit learning optimizer with black hole algorithm”, Electr. Power Syst. Res., Vol.
commitment via mixed integer linear programming: a case study of 108, pp. 16–34, 2014.
the three gorges project China”, IEEE Trans. Power Syst., Vol. 29, [24] Ignacio E. Grossmann, DICOPT user manual. Chemical Engineering Faculty,
No. 3, pp. 1232–1241, 2014. Carnegie Mellon University, Pittsburg, USA, 2018.
[7] Viana A. and Pedroso J. P., “A new MILP-based approach for unit [25] Sichard E. and Rosenthal, “GAMS, A User’s Guide”, Tutorial GAMS
commitment in power production planning”, Int. J. Electr. Power Development Corporation, Washington,2010.
Energy Syst., Vol. 44, pp. 997–1005, 2013

Authorized licensed use limited to: Murdoch University. Downloaded on June 14,2020 at 00:59:32 UTC from IEEE Xplore. Restrictions apply.

You might also like