Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Energy 251 (2022) 123905

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Quasi-oppositional turbulent water flow-based optimization for


cascaded short term hydrothermal scheduling with valve-point effects
and multiple fuels
V.P. Sakthivel a, *, K. Thirumal b, P.D. Sathya c
a
Department of Electrical and Electronics Engineering, Government College of Engineering, Dharmapuri, 636704, Tamilnadu, India
b
Department of Electrical and Electronics Engineering, Government College of Engineering, Bargur, 635104, Tamilnadu, India
c
Department of Electronics and Communication Engineering, Faculty of Engineering and Technology, Annamalai University, Chidambaram, 608002,
Tamilnadu, India

a r t i c l e i n f o a b s t r a c t

Article history: In an interconnected power system, short-term hydrothermal scheduling (SHTS) portrays a paramount
Received 1 October 2021 important task in the economic operation of electric power system. SHTS is discerned as a formidable
Received in revised form non-convex constrained optimization problem considering the cascade nature of hydro plants, and
28 January 2022
valve-point loading (VPL) and multi-fuel sources of thermal plants. In the present study, a novel met-
Accepted 1 April 2022
Available online 11 April 2022
aheuristic approach, a quasi-oppositional turbulent water flow-based optimization (QOTWFO) is pro-
posed to solve the SHTS problem. Turbulent water flow-based optimization (TWFO) is evolved by the
whirlpool prodigy formed in a turbulent flow of water. QOTWFO is an amended version of TWFO, where
Keywords:
Multi-fuel source
the quasi-opposition-based learning is pioneered for population initialization and iteration vaulting. It
Oppositional based learning generates a quasi-oppositional solution which has the highest probability to elicit the global optimum
Short-term hydrothermal scheduling solution than a randomly generated solution. Thus, the global searching ability and computational ef-
Turbulent water flow-based optimization ficiency of the algorithm is enhanced. A new heuristic constraints handling mechanism is proposed to
Valve point loading fulfil the equality constraints notably active power balance and dynamic water flow balance. The cogency
of proposed QOTWFO is examined on standard benchmark functions, and multi-reservoir cascaded
hydrothermal test systems with the cogitation of VPL effects and multi-fuel sources of thermal plants.
Besides, Newton Raphson load flow approach is employed to determine power losses and line flows in a
16-bus, and 35 transmission lines hydrothermal power system. To model a realistic power system model,
the maximum power transfer capability of each transmission line is taken into account. Simulation
outcomes substantiate that, compared with state-of-the-art heuristic techniques, QOTWFO offers better
feasible solutions as well as guarantees the robustness of the algorithm.
© 2022 Elsevier Ltd. All rights reserved.

1. Introduction reservoir, dynamic water flow balance and active power balance.
The limits of water discharge rate, reservoir storage volumes and
The reliability and security of power systems can be acquired by hydrothermal power generations are contemplated as inequality
the hydrothermal coordination. Short-term hydro-thermal sched- constraints. In addition to these constraints, valve point loading
uling (SHTS) is one of the main issues in economic operation of (VPL) effect is considered in the SHTS which devises a complex
power systems. It aims to determine the optimal power generation non-convex constrained optimization problem.
of hydro and thermal power plants by lessening the thermal fuel In the literature, a wide assortment of optimization approaches
cost in accordance with various equality and inequality constraints. has been developed to solve the SHTS problem which can be
The equality constraints comprise initial and final volumes of categorized into conventional optimization and heuristic tech-
niques. Salam et al. [1] proposed an improved Lagrangian relaxa-
tion based hydrothermal coordination algorithm, where the SHTS
* Corresponding author.
problem was segregated into thermal and hydro subproblems. Yang
E-mail addresses: vpsarvesh2013@gmail.com (V.P. Sakthivel), eeethirumal@ and Chen [2] developed multi-pass dynamic programming with
gmail.com (K. Thirumal), pd.sathya@yahoo.in (P.D. Sathya).

https://doi.org/10.1016/j.energy.2022.123905
0360-5442/© 2022 Elsevier Ltd. All rights reserved.
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Nomenclature URðiÞ; DRðiÞ ramp-up and ramp-down bounds of the ith thermal
unit, respectively
F total fuel cost Nl number of transmission lines
fi ðPs ði; tÞÞ thermal fuel cost function of ith plant in time period t Pline; Nl active power flow on transmission line Nl
Ps ði; tÞ power generation of ith thermal plant in time period max
Pline; maximum active power flow on transmission line Nl
Nl
t Nwh number of whirlpool sets
ai ; bi ; ci thermal fuel cost coefficients of the ith plant Np number of objects
di ; ei VPL effects coefficients of the ith thermal plant D number of decision variables
Ps L ðiÞ Ps U ðiÞ lower and upper thermal power limits of the rand1 ; rand2 random values between 0 and 1
ith plant di ; dj angles of the ith object and jth whirlpool respectively
Ns total thermal plants Dt weighted distance
T total intervals over a scheduling horizon Xi position of the ith object
aik ; bik ; cik ; dik ; eik thermal fuel cost coefficients of the ith plant randð1; DÞ D uniform randomly numbers generated within the
for the kth fuel source range [0, 1]
Cj1 ; Cj2 ; Cj3 ; Cj4 ; Cj5 ; Cj6 Generation coefficients of the jth hydro Whf and Whf the whirlpools with minimum and
plant maximum values of Dt , respectively
Nh number of hydro plants Ximin ; Ximax minimum and maximum bounds of ith object
Ph ðj; tÞ hydro power generation of jth plant in time period t FEi centrifugal force of ith whirlpool
Vh ðj; tÞ reservoir storage volume of the jth reservoir in time Whj position of the jth whirlpool
period t DWhj step size between the nearest and jth whirlpools
Qh ðj; tÞ water discharge rate of the jth reservoir in time f ðWhj Þ cost value of jth whirlpool
period t Xj;pbest personal best position of jth whirlpool
PD ðtÞ power demand in time period t f ðXj;pbest Þ cost value of jth whirlpool corresponds to its personal
PL ðtÞ transmission loss in time period t best position
Bij ; Bi0 ; B00 transmission loss coefficients Vpmin ; Vpmax maximum and minimum values of the vaulting
Ih ðj; tÞ; Sh ðj; tÞ inflow and spillage of jth hydro plant in time probability, respectively
period t iter current iteration number
tkj time delay between jth hydro plant and its kth itermax maximum number of generations
upstream plant Fmax objective function of the worst feasible solution in
Ruj number of upstream plants instantly above jth hydro the population
plant ε maximum absolute allowable constraint violation
Vh L ðjÞ; Vh U ðjÞ lower and upper water storage volume limits of CVðXÞ summation of constraint violation of solution X
the jth reservoir Wþ sum of the positive ranks for the problems wherein
Qh L ðjÞ; Qh U ðjÞ lower and upper water discharge limits of the jth the first algorithm excelled
reservoir W sum of the negative ranks
Vh ðjÞbegin ; Vh ðjÞend beginning and end reservoir storage volumes Ft average fuel cost obtained for the tuned parameter
of the jth hydro plant set
ndj number of prohibited discharge zones of jth hydro Fp average fuel cost obtained for the perturbed
plant parameter
Qh LB;n ðjÞ; Qh UB;n ðjÞ lower and upper discharge bounds of the nth
prohibited discharge zones of jth hydro
plant respectively

successive approximations to solve the daily hydrothermal coor- the earliest solutions.
dination problem in which computation time and memory were To defeat this lack, many researchers have focused on different
significantly reduced. In Ref. [3], a direct solution approach, semi- heuristic approaches. Orero and Irving [8] used genetic algorithm
definite programming was presented for the SHTS problem. Franco (GA) to the problem of estimating the optimal hourly plan of power
et al. [4] applied a network flow programming technique in hydro- generation in a hydrothermal system. The control parameters
dominated power system. A penalty approach was used to couple which pretend the behaviour of genetic algorithm were analysed. In
the hydro and electric variables. In Ref. [5], a mixed-integer model Ref. [9], an interactive fuzzy based evolutionary programming al-
was developed for short term hydrothermal planning to evade the gorithm was applied for short-term multi-objective hydrothermal
dilemmas provoked by non-linearity and non-convexity of the scheduling in which the non-smooth fuel cost and emission func-
problem. A compact mixed-integer linear programming technique tions of thermal power systems were considered. A variety of fast
on the basis of an equivalent unit model and a linear generation evolutionary programming (EP) approaches with Gaussian and
function was proposed by Guedes et al. [6]. other mutations was developed in Ref. [10] for the solution of SHTS
Despite the fact that these conventional approaches are direct problem and examined on a cascaded multi-reservoir hydropower
and proficient, yet have restricted capacity to address the SHTS system considering the VPL effects and prohibited operating zones
problem with non-smooth fuel cost function due to VPL effects and (POZ). In Ref. [11], two competing objectives which include fuel cost
non-convex hydro generation function [7]. These approaches and pollutant gas emissions were considered to solve the SHTS
depend on a few surmised models to solve the SHTS problem which problem. In this reference, the authors employed multi-objective
will definitely prompt to lower solution quality and also delicate to DE method to address economic environmental dispatch of

2
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

hydrothermal system. The SHTS problem was solved by differential handling method. In RCGA, simulated binary crossover and poly-
evolution (DE) algorithm in Ref. [12]. The superiority of DE algo- nomial mutation were employed in opposition with the single
rithm was revealed on two test systems and compared with erst- point crossover and bit-flipping mutation in BCGA. In Ref. [26], the
while approaches. SHTS problem was solved by fusing the RCGA and artificial fish
In [13], the authors implemented a modified DE (MDE) algo- swarm algorithm (AFSA) in which RCGA and AFSA were used as
rithm for attaining the optimal solution of economic load dispatch global and local searches respectively to improve the exploration
considering VPL effect. The mutation and selection mechanisms of and exploitation capabilities of the algorithm.
GA, DE and particle swarm optimization (PSO) were consolidated to Scheduling problem of hydrothermal units considering hy-
make another effective stochastic algorithm. Few PSO techniques draulic and electrical constraints was solved by RCGA with an
were employed for the solution of SHTS problem [14,15]. The improved Mühlenbein mutation (RCGA-IMM) in Ref. [27] to ensure
execution of PSO algorithms proved that the local variants of PSO the efficacy of the approach. The initialization and mutation steps
were appropriate to solve the SHTS problem. Mandal and Chakra- of DE were modified in Refs. [28,29] to manage the reservoir end
borty [16] extended the PSO algorithm to solve combined economic volume and the equality constraints effectively with a lesser
emission scheduling of hydrothermal plants with cascaded reser- computational effort. In Ref. [30], an improved PSO was presented
voirs wherein the VPL effects were engaged into consideration. In to the solution of SHTS problem, where dynamic search-space
Ref. [17], an effectual optimization using clonal selection algorithm squeezing strategy was adopted to handle the inequality con-
(CSA) was aimed at solving the SHTS problem. Simulation outcomes straints. Sakthivel et al. [31] applied a new multi-objective squirrel
proved that the CSA was competent in offering the global optimal search algorithm to solve the combined economic and environ-
solutions. mental power dispatch problem, where an external elitist de-
Roy [18] presented teaching learning-based optimization (TLBO) pository mechanism with crowding distance sorting was used to
to solve SHTS problem in which nonlinearities such as VPL effects of maintain the distribution diversity of Pareto-optimal solutions as
the thermal plant and prohibited discharge zones (PDZ) of water the evolution proceeds. In Ref. [32], a hybrid approach that
reservoir of the hydro plants were addressed. In Ref. [19], a nature comprised chaotic local search technique, grey wolf optimizer and
motivated heuristic algorithm, called as ant lion optimization (ALO) dragonfly approach was developed to solve the SHTS problem. The
was employed for a solution of practical SHTS problem with wind grey wolf optimization (GWO) and dragonfly algorithm (DA) were
integration. The random walk, roulette wheel and elitism mecha- utilized to enhance the local search ability with slow convergence
nisms were employed in ALO to increase exploration and exploi- and the global search capability with fast convergence respectively.
tation capabilities respectively. To enhance the diversity of population, a mutation manipulator
The main difference with the conventional strategies is that the was employed in the small population-based PSO (SPPSO)
heuristic strategies do not rely upon the mathematic model. approach for solving the SHTS problem in Ref. [33]. Besides, a DE
Nonetheless, rather than giving the global solution, several single algorithm was used to speed up the convergence of the algorithm.
heuristic approaches might deliver suboptimal solutions in solving In Ref. [34], daily hydrothermal generation scheduling was solved
nonconvex SHTS optimization problems. Furthermore, the heuristic by using modified adaptive PSO (MAPSO) in which the PSO co-
approaches such as GA and PSO experience the ominous effect of efficients were adaptably modified, and a novel velocity modifier
premature and slow convergences, which entail from a deficiency was introduced to increase the global search capability of the
of population diversity. Consequently, numerous improved and approach. Authors in Ref. [35] developed a modified social group
hybrid approaches were developed for solving the SHTS problem to optimization (MSGO), where a self-awareness probability factor
overwhelm the limitation of single heuristic approach. was embedded in the acquiring stage to improve both the explo-
A group search optimizer (GSO) was aimed to solve a reliability ration and exploitation aptitudes. In Ref. [36], a modified dynamic
constrained hydrothermal unit commitment problem in Ref. [20] to neighbourhood learning based PSO (MDNLPSO) in which the in-
lessen both the fuel and start-up costs of thermal power plants. formation was exchanged among the particles in the swarm was
Besides, a self-learning GSO (SLGSO) was developed by employing proposed to solve the SHTS problem. An optimal daily hydro gen-
adaptive covariance matrix and Le vy flights to rise the searching eration scheduling problem was solved by an enhanced PSO (EPSO)
ability and diversity of the algorithm respectively. Authors in in Ref. [37]. The improvements such as repellent concept and lo-
Ref. [21] employed krill herd algorithm (KHA) to address the SHTS gistic map based chaotic sequences were included in the approach.
problem in which the aptitudes of DE were adopted to improve the The SHTS problem was solved in Ref. [38] by adopting a couple-
convergence behaviour. In Ref. [22], crisscross search optimization based PSO (CPSO) to overwhelm the premature convergence
(CSO) was developed to solve the practical SHTS problem by dilemma. To weigh the global and local search abilities of sine
employing horizontal and vertical crossover operators. The global cosine algorithm (SCA), in Ref. [39], the control parameters were
search ability of the approach was improved by cross-border search properly selected. Narang [40] presented an improved predator
concept, whereas the local search ability was addressed by adopt- influenced civilized swarm optimization (IPCSO) approach for the
ing an arithmetic crossover operator. In Ref. [23], an efficient dif- solution of practical SHTS and profit-based SHTS problems in light
ferential real-coded quantum-motivated evolutionary algorithm of VPL effects and PDZ on reservoir discharge. By developing a new
(DRQEA) to solve the SHTS problem was employed. To enhance the hybrid algorithm, in Ref. [41], chemical reaction optimization and
global searching capability, real-coded rule, and adaptive mutation DE techniques were put forward to solve the combined economic
and crossover operators were incorporated in DRQEA. A clonal real- emission SHTS problem.
coded quantum-inspired evolutionary algorithm (CRQEA) with The optimal solution of SHTS problem was attained in Ref. [42]
Cauchy mutation that avoids premature convergence was devel- by utilizing an adaptive chaotic artificial bee colony algorithm. A
oped and aimed to solve the SHGS problem in Ref. [24]. To evade combination of chaotic search and an adaptive coordinating
the premature convergence of CRQEA, the clonal operator and mechanism was employed on the optimization problem to evade
Cauchy mutation were embedded. from a local optimum and to avoid premature convergence. In
Binary coded GA (BCGA) and real coded GA (RCGA) were pro- Ref. [43], Quasi-oppositional group search optimization (QOGSO)
posed in Ref. [25] for the solution of SHTS problem considering which mimicked quasi-oppositional based learning was presented
continuous and non-convex fuel cost function. Both the approaches to improve the efficacy and solution quality. A disruption based
were incorporated with an efficient parameter less constraint- gravitational search algorithm (DGSA) to ascertain the solution for
3
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

SHTS problem was developed in Ref. [44]. In this technique, a populace to accomplish the best population diversity compared
disruption manipulator subject to astrophysics was unified into with any other metaheuristics.
gravitational search algorithm (GSA) to boost its operation. In this study, the solution of SHTS with cascaded reservoir
In [45], a hybrid algorithm founded on teaching learning-based problem is performed by using quasi-oppositional turbulent water
optimization (TLBO) and oppositional based learning (OBL) was flow-based optimization (QOTWFO) which is developed as a new
introduced for solving the SHTS problem. Basu [46] introduced an approach. The main contributions of this study are summarized
improved DE (IDE) in which the scaling factor of DE was replaced by hereunder:
Gaussian random variable to enhance the searching ability and
employed to solve the SHTS problem. Authors in Ref. [47] applied a  To the best of authors' knowledge, this is the first time that the
real-coded chemical reaction optimization (RCCRO) for solving the TWFO algorithm is employed to solve the SHTS with cascaded
problem. Furthermore, the same authors developed an opposi- reservoir problem.
tional based RCCRO (ORRCCRO) in which opposition-based learning  A novel TWFO algorithm is developed, where a quasi-
was employed for initialization of population to generate high- oppositional-based learning is employed into the TWFO algo-
quality solutions [48]. rithm to ameliorate the performance of the algorithm.
In [49], quasi-reflected ions motion optimization (QRIMO) for  The effect of VPL, and multiple fuel sources of thermal plants,
solving the SHTS problem was described. Two stages such as liquid prohibited discharge zones (PDZ) of hydro plants, and the
and crystal stages were employed to balance the global and local transmission losses of hydrothermal plants are addressed for
searches of the algorithm. Besides, quasi-reflected-based learning obtaining the optimal solution of the SHTS problem.
was adopted to ions motility to improve the convergence rate and  Newton Raphson load flow approach is employed to determine
solution quality. The quasi-reflected strategy was coupled into line flows and power losses in a lossy hydrothermal power
symbiotic organisms’ search (SOS) in Ref. [50] to enhance the so- system by taking into account the transmission line security
lution space of algorithm. constraints.
Hybrid approaches are able to leverage the benefits of inte-  The superiority of the suggested QOTWFO approach is exam-
grated heuristic approaches and consequently improve the exhi- ined on benchmark functions and six SHTS test cases, and also
bition in intermingling speed and solution quality. Nevertheless, it compared with the erstwhile approaches in the literature.
stays a grave challenge to equipoise the exploration and exploita-
tion searching capability of these approaches in solving nonconvex The rest of the paper is systematized as follows: Section 2 pre-
large-scale SHTS problem. They are difficult to yield global optimal sents the problem formulation of SHTS model. An overview of
solution and do not assure the convergence in limited time. TWFO algorithm is described in Section 3. Sections 4 and 5
Consequently, it remains a substantive issue for heuristic ap- respectively introduce the quasi opposition-based learning strat-
proaches to defeat the untimely union and to enhance the solution egy and the proposed QOTWFO algorithm. The constraints
quality of large-scale SHTS issue. handling strategy and the implementation of QOTWFO algorithm
Recently, a new optimization algorithm named as turbulent to the solution of SHTS problem are explained in Section 6. Results
water flow-based optimization (TWFO) proposed by Ghasemi et al. attained by the suggested approach and the erstwhile approaches
has become a contender for optimization applications because of its are discussed in Section 7. Finally, the conclusions are epitomised in
adaptability and proficiency [51]. TWFO is a metaheuristic algo- Section 8.
rithm that mimics the demeanour of whirlpools formed in rivers,
seas, and oceans due to turbulent flow of water. This algorithm has 2. Problem formulation of SHTS model
outclassed DE, PSO, GWO, artificial bee colony (ABC) and
biogeography-based optimization (BBO) on solving different In SHTS problem, as the hydropower generation cost is frivolous
benchmark test functions and economic load dispatch problems. In than that of thermal plants, the aim is to determine the optimal
comparison with other metaheuristic approaches, TWFO compre- water discharges of hydro reservoirs and thermal power generation
hends the following advantages for offering the globally optimal by minimizing the total thermal generation cost over a 24 h’ time
solution of problem: horizon with the satisfaction of different equality and inequality
constraints. The objective function and constraints of SHTS problem
 TWFO algorithm has no algorithm specific parameters which are defined in the following manner:
results in enhanced convergence rate over other metaheuristics.
 TWFO employs weighted distance measure among the whirl-
2.1. Objective function
pools in its optimization process for better exploration.
 Furthermore, two distinct interactions such as object's interac-
The objective function of SHTS problem for Ns thermal power
tion of its set and other whirlpools, and interactions among the
plants with T time span can be defined as follows:
whirlpools are incorporated in TWFO which preclude the
approach from getting stuck at a local optimum. X
T X
Ns
 The integration of centrifugal force phase enhances the popu- Min F ¼ fi ðPs ði; tÞÞ (1)
lation diversity of TWFO algorithm. t¼1 i¼1

The quadratic fuel cost function of ith thermal plant is described


Nevertheless, for large-scale highly constrained non-convex
by
optimization problems, TWFO experiences slow convergence
speed and inferior exploration ability as like other metaheuristics.
fi ðPs ði; tÞÞ ¼ ai þ bi Ps ði; tÞ þ ci ðPs ði; tÞÞ2 (2)
Thus, in this paper, a machine learning strategy, quasi-oppositional
based learning (QOBL) is employed in the population initialization In modern thermal power plants, multi-valve turbines are pro-
and generation vaulting phases. At the end of each iteration, the vided to deliver required incremental power. The VPL effect is
feasible solutions acquired by QOBL approach are included into the emerged by steam valve opening in a turbine, which devises the
optimization process by excluding the inferior quality solutions fuel cost function as nonconvex. When the VPL effects on thermal
from the current population of whirlpools. This empowers the power plants are considered, the rectified sine component is
4
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

appended in the quadratic objective function, as listed in the scheduling horizon are fixed by mid-term scheduling processes,
following equation: and are expressed as follows:

fi ðPs ði; tÞÞ ¼ ai þ bi Ps ði; tÞ þ ci ðPs ði; tÞÞ2 Vh ðj; tÞjt¼0 ¼ Vh ðjÞbegin (9)
þ jdi sinðei  ðPs L ðiÞ  Ps ði; tÞÞÞj (3)

In addition to the VPL effects, multiple fuel sources of thermal


Vh ðj; tÞjt¼T ¼ Vh ðjÞend (10)
power plants are engaged into account and can be expressed as
under:

8
>
> Fuel type 1; ai1 þ bi1 Ps ði; tÞ þ ci1 ðPs ði; tÞÞ2 þ jdi1  sinðei1  ðPs L ðiÞ  Ps ði; tÞÞÞj
<
fi ðPs ði; tÞÞ ¼ Fuel type 2; ai2 þ bi2 Ps ði; tÞ þ ci2 ðPs ði; tÞÞ2 þ jdi2  sinðei2  ðPs L ðiÞ  Ps ði; tÞÞÞj (4)
>
>: : : :
:
Fuel type k; aik þ bik Ps ði; tÞ þ cik ðPs ði; tÞÞ2 þ jdik  sinðeik  ðPs L ðiÞ  Ps ði; tÞÞÞj

2.3. Inequality constraints

The hydro power generation is pondered to be a polynomial 2.3.1. Power generation limits
function of turbine discharge rate and reservoir storage volume. It The power generation of each hydro and thermal plants at each
can be defined as follows: time horizon shall reside within its lower and upper limits, and is
expressed as follows:
Ph ðj; tÞ ¼ Cj1 ðVh ðj; tÞÞ2 þ Cj2 ðQh ðj; tÞÞ2 þ Cj3 Vh ðj; tÞQh ðj; tÞ
(5) Ps L ðiÞ  Ps ði; tÞ  Ps U ðiÞ; i2 Ns (11)
þ Cj4 Vh ðj; tÞ þ Cj5 Qh ðj; tÞ þ Cj6

Ph L ðjÞ  Ph ðj; tÞ  Ph U ðjÞ; j2 Nh (12)

2.2. Equality constraints

2.2.1. Real power balance 2.3.2. Ramp rate limits of thermal power plants
At each time horizon, the total power generation of hydro and The power generation of each thermal plants can be increased
thermal plants ought to meet the load demand and transmission or decreased within its given limits, which can be defined by the
losses, and can be expressed as follows: following inequality constraints:

Ps ði; tÞ  Ps ði; t  1Þ  URðiÞ; i2 Ns (13)


X
Ns X
Nh
Ps ði; tÞ þ Ph ðj; tÞ ¼ PD ðtÞ þ PL ðtÞ; i 2 Ns ; j 2 Nh ; t2T
i¼1 j¼1 Ps ði; t  1Þ  Ps ðtÞ  DRðiÞ; i 2 Ns ; t2T (14)
(6)
The transmission losses can be determined by Kron's loss for-
mula as follows: 2.3.3. Active power flow limits
The power transfer capability of each transmission line is given
h þNs NX
NX h þNs h þNs
NX
as:
PL ðtÞ ¼ Pði; tÞBij Pðj; tÞ þ Bi0 Pði; tÞ þ B00 (7)
 
i¼1 j¼1 i¼1 P   P max (15)
line; Nl line; Nl

2.2.2. Reservoir flow balance


2.3.4. Reservoir storage volume constraints
The storage of hydro reservoirs relies on reservoir storage in
The operating reservoir volume in each hydro power plant at
preceding span, inflow, water supply from upstream reservoir and
each time interval have to be limited between its lower and upper
spillage. In this work, the water discharge lag between reservoirs is
limits, which can be defined below:
taken into account. It can be formulated as follows:
Vh L ðjÞ  Vh ðj; tÞ  Vh U ðjÞ; j 2 Nh ; t2T (16)
Vh ðj; tÞ ¼ Vh ðj; t  1Þ þ Ih ðj; tÞ  Qh ðj; tÞ  Sh ðj; tÞ
Ruj 
X    
þ Qh k; t  tkj þ Sh k; t  tkj (8)
k¼1 2.3.5. Water discharge rate constraints
The water discharge rate of each hydro plant must be bounded
to its lower and upper values, which can be defined as:

2.2.3. Initial and terminal reservoir storage volumes Qh L ðjÞ  Qh ðj; tÞ  Qh U ðjÞ; j 2 Nh ; t2T (17)
The initial and final storage of reservoirs over the whole
5
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 1. Population of whirlpools.

3. Turbulent water flow-based optimization

Turbulent Water Flow-Based Optimization (TWFO) is one of the


recent metaheuristics which is developed by Ghasemi et al. on the
metaphor of whirlpool prodigy due to turbulent flow of water in a
river, sea or ocean [51]. When water passes through a narrow
opening by gravity, it starts to spin and forms a cavity in the focal
point of the channel. Subsequently, it can mature into a complete
whirlpool. The cavity makes a vacuum into which objects like air
pockets, water particles, and other gliding objects are pulled down.
As these objects are sucked into the vortex, the centrifugal force
preserves the hole in the centre through which air passes. The
centripetal force makes an object to move in a circular motion with
uniform speed.
TWFO algorithm is a population-based approach consisting of
several whirlpools or groups in which each object explores through
a multi-dimensional search space to offer the global optimum so-
lution. The object closer to the hole of whirlpool is considered as
the best member of the group, which ingests the objects around
them in accordance with their distances and locates them squarely
at the middle. This implied that whirlpool imposes centripetal force
on all objects and particles in the surrounding area. The centripetal
force is the force acting towards the centre of circular path and is
expressed as follows:
Fig. 2. Concept of position modification of an object.
mv2
Fc ¼ (19)
r
The water discharge rate of the hydro plant reservoirs with
prohibited discharge zones (PDZ) can be modelled as follows: where, m and v are the mass and velocity of object, and r is the
radius of the circle. Moreover, the mass, centrifugal force, and as-
8 sociation between nearby whirlpools and the objects cause oppo-
< Qh L ðjÞ  Qh ðj; tÞ  Qh LB;1 ðjÞ
Q ðjÞ  Qh ðj; tÞ  Qh LB;n ðjÞ n ¼ 2; 3; : : ; ndj sition and deviation in the objects of all whirlpools.
(18)
: Qh UB; n1 The whirlpools around the objects move the best individual of
h UB; n ðjÞ  Qh ðj; tÞ  Qh U ðjÞ
the group in the opposite direction and lessen the centripetal
forces. The centrifugal force might cause an antagonistic impact on

6
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

this centripetal force and change the position of the objects. In assumed to be located at the centre of the whirlpool which is
addition, greater and more powerful whirlpools pull or push the depicted in Fig. 1.
more modest and vulnerable ones as indicated by the direction and
distance, and modify the positions of other whirlpools. 3.2. Impacts of whirlpools on objects of its set and different
In this optimization algorithm, the positions of objects in the whirlpools
whirlpools are considered as different decision variables, and the
distance between the best member and members in a group is Each whirlpool acts like a sucking hole and gravitates to merge
analogous to the fitness value of the objective function. In TWFO, the positions of objects in its set through its mid-position aided by a
each object in the whirlpool alters its position and moves to a better centripetal force acting upon them. However, the positions of ob-
position as iteration proceeds. The virtual model of objects inter- jects can be changed by the distance between the whirlpools. Each
action between the whirlpools is a reiterative process, which object moves from the current position to the next one by a step
comprises five phases as follows: size, Dxi with an angle, d near its whirlpool's centre as shown in
Fig. 2. The angle between the whirlpool's hole and the object is
 Development of whirlpool sets updated in each iteration as follows:
 Impacts of whirlpools on objects of its set and different
whirlpools dnew
i ¼ di þ rand1  rand2  p (20)
Centrifugal force
The whirlpools with the highest and the lowest weighted dis-
Whirlpools' interactions
tance amongst the objects are determined using the following
 Survival of the fittest member in each whirlpool's set
equation:
The description and mathematical model of these phases are
Dt ¼ f ðWht Þ  jWht  sumðXi Þj0:5 (21)
given as follows:
The step size of the Dth component of the ith object may be
modified by
3.1. Development of whirlpool sets
      
DXi ¼ cos dnew
i  randð1; DÞ  Whf  Xi  sin dnew
i
The initial population of the algorithm which comprises Np       
members is equally distributed between Nwh whirlpool sets or  randð1; DÞ ðWhw  Xi Þ  1 þ cos dnew
i sin dnew
i

groups. Each member represents a candidate solution and may be
designated a fitness value. The best member of each whirlpool set is (22)

Fig. 3. Illustration of QOBL strategy.

7
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 4. Pseudocode for the proposed QOTWFO algorithm.

The position update of the jth whirlpool at ith object may be    


expressed by Xinew ¼ min max Xinew ; Ximin ; Ximax (24)

Xinew ¼ Whj þ DXi (23)


3.3. Centrifugal force
The updated position of an object is not constantly ensured to
meet out the inequality constraints on account of over/under In virtue of Newton's first law, an object at rest or uniform
swiftness. If any object violates its inequality constraint, then the motion will persist at rest or uniform motion unless a net external
position of the object is fixed to its upper/lower limit. Consequently, force acts on it. Sporadically, centrifugal force whelms the cen-
this can be modelled as follows: tripetal force of the whirlpool and arbitrarily moves the object to
8
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

another position. To model this phase, the centrifugal force in solution is routinely nearer than an opposite candidate solution to
accordance with the angle between the object and the whirlpool is the global best solution [43]. To enhance the computation efficiency
determined by and solution quality, quasi-opposition-based learning (QOBL) has
employed in the TWFO algorithm. The QOBL is adopted in popu-
  new 2   new 2
FEi ¼ cos di  sin di (25) lation initialization and iteration vaulting.
In a d-dimensional search space, if pðX1 ; X2 ; : : : Xi : : :XD Þ is a
If the value of centrifugal force for ith object is more than a point within the range ðXimin ; Ximax Þ, its opposite and quasi-
random value, the position of a randomly selected pth element of opposite points are defined in Eqs. (31) and (32) respectively as
the object is reinitialized as follows: follows:
  h i
xi;p ¼ xmin
p þ rand  xmax
p  xmin
p (26) Xio ¼ Ximin þ Ximax  Xi ; Xi 2 Ximin ; Ximax ; i ¼ 1; 2; : : D
(31)

 
3.4. Whirlpools’ interactions Xiqo ¼ rand ci ; Xio ; i ¼ 1; 2; : : D (32)

There is a tendency of each whirlpool to attract the remaining where, ci is the centre of search space and can be defined by:
whirlpools by applying centripetal force to submerge them into
their respective holes. To model this effect, the whirlpool with the Ximin þ Ximax
ci ¼ (33)
least weighed distance from all the whirlpools is calculated as 2
follows: As a function of vaulting probability (Vp ), the TWFO algorithm is
   contrived to provide a new solution which is better than the current
Dt ¼ f ðWht Þ  Wht  sum Whj 0:5 (27) best solution. The value of vaulting probability is typically chosen as
The step size between the least weighted distance whirlpool ½0; 0:6 and determined as:
and the jth whirlpool is expressed as follows:     iter
       Vp ¼ Vpmax  Vpmin  Vpmax  Vpmin  (34)
DWhj ¼ randð1; DÞ  cos dnew  itermax
j þ sin dnew
j  Whf  Whj 
After generating the new whirlpool sets, the quasi-oppositional
(28) whirlpool sets are generated based on the vaulting probability.
The jth whirlpool's position can be updated in accordance with Then, the Np fittest members are selected from the union of the
Eq. (29). new whirlpool and the quasi-oppositional whirlpool sets, and used
them for the next iteration. The illustration of QOBL strategy is
Whnew
j ¼ Whf þ DWhj (29) shown in Fig. 3.

5. Quasi-opposition turbulent water flow optimization


3.5. Survival of the fittest member in each whirlpool's set
In this paper, the authors have proposed a quasi-opposition
If the new objective value for any whirlpool is better than its turbulent water flow optimization (QOTWFO)algorithm in which
preceding value, the new position of that whirlpool can be modified
as follows:
(    
Xj; pbest ; if f Xj;pbest < f Whj
Whj ¼ (30)
Whj ; otherwise

The best member among the whirlpools is determined as the


global best solution.

4. Quasi opposition-based learning

In Ref. [52], Tizhoosh developed an efficient learning strategy,


opposition-based learning (OBL) to enhance the computational
efficiency of various heuristic algorithms. In the OBL method, cur-
rent population and its opposite population are pondered to get
better solution. It has been demonstrated in the literature that an
opposition candidate solution has a greater likelihood to advance
towards the global optimum solution over a random candidate
solution. Accordingly, as compared a solution to its opposite, a
Fig. 5. Pseudocode for handling the reservoir flow balance constraint.
lesser search region is expected for converging to the best solution.
Furthermore, it has emerged that a quasi-opposite candidate

9
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 6. Pseudocode for handling the reservoir storage limits.

QOBL theory is employed in the original TWFO algorithm to 6. Constraints handling mechanism
quicken the convergence mobility and to find the global optimal
solution. The QOTWFO algorithm engages the QOBL in population 6.1. Inequality constraints handling rule
initialization and iteration vaulting. The search process of the
proposed QOTWFO algorithm is depicted in Fig. 4. The inequality constraints such as water discharge limits,
reservoir storage limits, and hydro and thermal generation limits
are taken into account.

10
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

X
T
DVh ðjÞ ¼ Vh ðjÞbegin  Vh ðjÞend þ ð Ih ðj; tÞ  Qh ðj; tÞ  Sh ðj; tÞÞþ
t¼1
T 
Ruj X
X    
Qh k; t  tkj þ Sh k; t  tkj
k¼1 t¼1
(41)
In this paper, a two-stage heuristic constraint handling strategy
is employed to equally distribute the total storage volume violation
at each time horizon, and then randomly different time horizons
are selected to modify the water discharge till the constraint is
repaired. The procedure for satisfying the reservoir flow balance
constraint of hydro plant is summarized in Fig. 5.

6.3. Reservoir storage limits constraints handling

The reservoir flow balance constraint is satisfied by employing


the heuristic constraint handling strategy which is summarized in
Fig. 7. Pseudocode for handling the power balance constraint.
Section 6.2. However, the reservoir storage volume limit might be
violated. The steps for repairing the reservoir storage volume
With a view to satisfy the water discharge constraint, the violation in hydrothermal scheduling are given in Fig. 6.
following equation is employed:

Qh ðj; tÞ ¼ minðmaxðQh ðj; tÞ; Qh L ðjÞÞ; Qh U ðjÞÞ (35) 6.4. Power balance constraint handling

Considering the PDZ, the handling method for water discharge To meet out the real power balance constraint, the thermal
constraint is modelled as follows: power generation is repaired without modifying the hydro power
8 generations in the preceding amendment. The violation of real
> Qh L ðjÞ; if Qh ðj; tÞ < Qh L ðjÞ
>
> power balance at the tth time horizon can be defined as follows:
>
> Q h ðj; tÞ; if Qh L ðjÞ  Qh ðj; tÞ  Qh U ðjÞ
>
>
>
< Qh U ðjÞ; if Qh ðj; tÞ > Qh U ðjÞ X
Ns X
Nh
Qh ðj; tÞ ¼ QhLB;n ðjÞ; if Qh LB; n ðjÞ < Qhðj; tÞ < Qh UB; n ðjÞ and DPs ðtÞ ¼ PD ðtÞ þ PL ðtÞ  Ps ði; tÞ  Ph ðj; tÞ (42)
>
>
>
> Qh ðj; tÞ  Qh LB;n ðjÞ < Qh UB; n ðjÞ  Qh ðj; tÞ i¼1 j¼1
>
>
>
> Qh LB;n ðjÞ; if Qh LB; n ðjÞ < Qh ðj; tÞ < Qh UB; n ðjÞ and
:     The technique for satisfying the power balance constraint which
Qh ðj; tÞ  Qh LB;n ðjÞ < Qh UB; n ðjÞ  Qh ðj; tÞ
is summarized in Fig. 7, has the imitative concept of reservoir flow
(36) balance constraint handling.
The handling method for thermal generation limits and ramp
rate limits can be defined in Eqs. 37e39.
6.5. Spillage modelling
Ps ði; tÞ ¼ minðmaxðPs ði; tÞ; Ps L ðiÞÞ; Ps U ðiÞÞ (37)
In this work, water spillages are considered to satisfy the lower
and upper generations limits of hydro plants. The water spillage can
Ps L ði; tÞ ¼ maxðPs L ðiÞ; Ps ði; t  1Þ  DRðiÞ Þ (38)
be modelled as follows:
8
Ps U ði; tÞ ¼ minðPs U ðiÞ; Ps ði; t  1Þ þ URðiÞ Þ (39) < Qh ðj; tÞ; Ph ðj; tÞ < Ph L ðjÞ
Sh ðj; tÞ ¼ 0; Ph L ðjÞ  Ph ðj; tÞ  Ph U ðjÞ (43)
:
Qh ðj; tÞ  Qh ðjÞjPh U ðjÞ; Ph ðj; tÞ > Ph U ðjÞ

6.2. Reservoir flow balance constraint handling where, Qh ðjÞjPh U ðjÞ is the water discharge of jth reservoir corre-
sponding to its maximum power generation.
The equality constraints of hydro power plant include the
reservoir flow balance and, initial and terminal reservoir storage
6.6. Selection strategy rooted in constraint violation
volumes. These equality constraints are satisfied by modifying the
storage volume and water discharges of hydro plants. The violation
The position of objects is updated by the above two-stage
of final reservoir storage volume can be defined as follows:
constraint handling mechanism. Nevertheless, the modified solu-
tions are infeasible because of the complex SHTS problem. Conse-
DVh ðjÞ ¼ Vh ðj; TÞ  Vh ðjÞend (40)
quently, the violation of each individual is determined on the basis
Based on Eqs. (9) and (10), the violation of the final volume for of reservoir storage limit and real power balance constraint as
the jth hydro plant is expressed as follows: follows:

11
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 8. Flowchart of QOTWFO algorithm for the solution of SHTS problem.

12
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 1 three different hydrothermal test systems for solving the SHTS
Benchmark test functions. problem with cascaded reservoirs. The performances of QOTWFO
Function Formulation Domain algorithm are compared with TWFO and other erstwhile tech-
P
niques in the literature. Due to stochastic nature of the QOTWFO
Sphere D ½100; 100D
f1 ðxÞ ¼ x2i algorithm, 50 independent runs are performed for each test case.
i¼1
Schwelfel's 2.22 PD QD ½10; 10D
The proposed algorithm is coded in MATLAB (R2019B) environment
f2 ðxÞ ¼ jxi j þ jxi j
i¼1
i¼1 and implemented on a PC with Intel Core-i5 2.4 GHz processor and
Quartic PD ½1:28; 1:28D 8 GB RAM.
f3 ðxÞ ¼ ix4 þ Rand
i¼1
Alpine PD ½10; 10D
f4 ðxÞ ¼ jxi sinxi þ 0:1xi j 7.1. Benchmark function evaluation
i¼1

In this study, 4 benchmark functions including three unimodal


and one multimodal functions with 30 dimensions are used to
Table 2
validate the robustness of the proposed QOTWFO approach. The
Statistical results for Benchmark test functions.
benchmark functions and their domains are given in Table 1. More
Function Function value Algorithm precisely, the population size, number of whirlpools and the
TWFO QOTWFO maximum number of iterations are respectively set as 100, 8 and
Sphere Minimum 2.5685Ee27 2.5685Ee27
500. Table 2 details the statistical analysis of the benchmark func-
Mean 4.3774Ee21 4.2685Ee23 tions, which are recorded after 50 autonomous trials. As can be
Maximum 3.2202Ee19 4.0957Ee20 seen from the table that the QOTWFO yields better solutions in
Standard deviation 3.9245Ee20 5.3281Ee22 terms of minimum, mean and maximum objective values that the
Schwelfel's 2.22 Minimum 6.5819Ee15 6.9543Ee16
TWFO and other state-of-the-art metaheuristics reported in
Mean 5.5448Ee12 4.6509Ee14
Maximum 1.8387Ee10 2.5863Ee11 Ref. [53]. In addition to the best solutions, QOTWFO excelled the
Standard deviation 2.7362Ee11 4.9426Ee14 other optimizers in the context of robustness, which reflects its
Quartic Minimum 2.4408E03 3.1375E04 incredible potential to solve the complex power system optimiza-
Mean 2.6113E-03 4.1953E-04 tion problems.
Maximum 3.5635E-03 5.1948E-04
Standard deviation 4.5538E-04 2.8469E-05
Alpine Minimum 1.4945E15 2.6931E16 7.2. Characterization of test cases
Mean 3.5692E11 3.1503E12
Maximum 1.6935E09 2.1976E09
The various attributes of the test systems include VPL, trans-
Standard deviation 2.4861E11 4.85291E14
mission losses, PDZ and multi-fuel source. In this work, the time lag
of water transfer among the consecutive reservoirs are being taken
into account. Additionally, the spillage effect is considered in the
reservoir flow balance to avoid the infeasible hydro power gener-
X
T X
Nh
CVðXÞ ¼ maxð0; Vh ðj; tÞ  Vh U ðjÞ; Vh L ðjÞ  Vh ðj; tÞÞþ ation. Based on the attributes of the test systems, six different test
t¼1 j¼1 cases with varying complexities are analysed. These test cases are
0  10:5 presented in Table 3. The framework of cascaded hydroelectric
X  X X 
T  Ns Nh  power generators for cases 1 to 4, case 5 and case 6 are depicted in
@ PD ðtÞ þ PL ðtÞ  Ps ði; tÞ  Ph ðj; tÞ A

t¼1  i¼1 j¼1  Figs. 9-11 respectively. The entire scheduling time is equally par-
titioned into 24 h’ time horizon with every hour a span, and in
(44) every span hydro and thermal power generations are firmly iden-
The fitness value of each object can be defined as: tified to accomplish the power balance constraint.

F; if X is fesible 7.3. Parameter tuning
fit ðFÞ ¼ (45)
Fmax þ CVðXÞ; otherwise
The parameters such as population size ðNp Þ, number of whirl-
The application of QOTWFO algorithm for the solution of SHTS
problem is displayed in Fig. 8. pools ðNwh Þ and maximum vaulting probability limit ( Vpmax ) have a
crucial impact on the performance of the algorithm. Consequently,
the calibration of these parameters is a legitimate challenge for the
7. Simulation results and analysis implementation of QOTWFO algorithm to offer the best feasible
solution for each test case of the SHTS problem. In this research
To authenticate the achievability and potency of the suggested work, Taguchi strategy is applied to assign the parameters of the
QOTWFO algorithm, it is examined on benchmark functions and algorithm. The primary benefit of this strategy is to lessen the

Table 3
Attributes of the examined test cases.

Case studies Ns Nh VPL Transmission losses PDZ Multi-fuel

Case 1 1 equivalent unit 4-unit 7 7 7 7


Case 2 1 equivalent unit 4-unit ✓ 7 ✓ 7
Case 3 3-unit 4-unit ✓ ✓ 7 7
Case 4 10-unit 4-unit ✓ 7 7 ✓
Case 5 5-unit 4-unit ✓ ✓ 7 7
Case 6 40-unit 10-unit ✓ 7 7 7

13
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 9. Hydro model with cascaded power plants for cases 1 to 4.

computation time with a reduced number of experiments. The


number of experiments required for selecting the three parameters,
each with 4 levels is 81 (34 Þ. Nevertheless, Taguchi strategy requires
only 16 experiments.
Each parameter has four levels as given in Table 4. L16 orthog-
onal array based on the Taguchi method is used to determine the
Fig. 10. Hydro model with cascaded power plants for case 5.
optimal parameters. Table 5 presents the value of tuned parame-
ters. The parameters are tuned at Run # 1 (a, b, c: 1, 1, 1) for Cases 1
and 2, Run # 6 (a, b, c: 2, 2, 1) for Case 3 and Run # 10 (a, b, c: 3, 2, 4) the suggested algorithm. As can be seen in Table 7 and Fig. 12 that
for Cases 4, 5 and 6 in the Taguchi orthogonal array. the power generated by the hydrothermal power plants gratify the
real power balance constraint which is defined in Eq. (6).
7.4. Case 1 The storage volume of all the reservoirs which is determined by
Eq. (8) is displayed in Fig. 13. It is noticed that the QOTWFO algo-
The first case examines a modest hydrothermal system rithm satisfies the initial and end reservoir limits and the reservoir
comprising four hydro plants and an analogous thermal plant. The storage volume limits. Table 8 presents the statistical results of fuel
VPL effects and transmission losses are ignored. The system data costs and computation time obtained by the QOTWFO, TWFO and
including inflow rate, storage volume limits, generation co- other erstwhile approaches. When Table 8 is analysed, it is note-
efficients, hydro and thermal power generations limits and load worthy that the QOTWFO attains the best findings with regard to
demand for each time span of this test case are adopted from minimum (922284.05 $/day), maximum (922659.69 $/day), and
Ref. [8]. The water inflow rate of each reservoir is kept constant over mean (922498.56 $/day) fuel costs among the TWFO and other
the entire time span. The thermal fuel cost coefficients a, b and c are compared approaches, verifying the best solution quality of
chosen as 5000, 19.2 and 0.002 respectively. The power generation QOTWFO algorithm. Moreover, the standard deviation obtained by
of thermal plant is limited between 500 and 2500 MW. The optimal the QOTWFO algorithm is the least, which proves the stability of
water discharges of hydro plants and thermal power output at each the algorithm.
time interval found by the QOTWFO algorithm are presented in
Tables 6 and 7 respectively. The power generation of each hydro
power plant is estimated according to Eq. (5). Fig. 12 shows the
hydro and thermal power output at each time horizon attained by

14
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 11. Hydro model with cascaded power plants for case 6.

Table 4
Calibration of QOTWFO.

Parameters Level 1 Level 2 Level 3 Level 4

a: Population size ðNp Þ 60 80 100 120


b: Number of whirlpools ðNwh Þ 4 8 12 16
c: Maximum vaulting probability limit ( Vpmax ) 0.3 0.4 0.5 0.6

7.5. Case 2 discharges of hydro plants and thermal power generation attained
by the proposed approach are reported in Tables 9 and 10 respec-
The second case considers the hydrothermal power system of tively. Also, the hourly hydro power generation is included in
Case 1 with PDZs of hydro plants and VPL effects of thermal power Table 8. Figs. 14 and 15 display the power generation of hydro and
plants. The coefficients related to the VPL of thermal plant, d and e thermal power plants, and reservoir storage volume during 24 h
are 700 and 0.085 respectively. Due to the consideration of VPL respectively. From Table 10 and Fig. 14, it is obvious that the optimal
effects, the SHTS problem becomes nonconvex and non- hydro and thermal power generation attained by the QOTWFO
differentiable, and it is relatively tough to solve. The optimal hourly algorithm convinces the power demand at each time horizon

15
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 5 without any constraint violation. It seems obvious from Fig. 15 that
Tuned parameters of QOTWFO. the proposed algorithm is satisfying all the reservoir storage vol-
Test cases Np Nwh Vpmax ume constraints.
The statistical results obtained by QOTWFO, TWFO and other
Case 1 60 4 0.3
Case 2 60 4 0.3 heuristic approaches are compared in Table 11. According to
Case 3 80 8 0.3 Table 11, the fuel cost and standard deviation obtained by QOTWFO
Case 4 100 8 0.6 are 924543.7617 $/day and 6.92, which are significantly lesser than
Case 5 100 8 0.6 those obtained by the compared algorithms. Thus, the QOTWFO
Case 6 100 8 0.6
algorithm epitomizes the better robustness among the compared

Table 6
Optimal hourly hydro discharge and hourly spillage attained with the proposed QOTWFO algorithm for case 1.

Hour (t) Hydro discharges (104 m3 ) Hydro spillages (104 m3 )

Qh ð1; tÞ Qh ð2; tÞ Qh ð3; tÞ Qh ð4; tÞ Sh ð1; tÞ Sh ð2; tÞ Sh ð3; tÞ Sh ð4; tÞ

1 10.000033 7.300246 0 13.000000 0 0 30.0000 0


2 10.500783 6.561310 0 13.000000 0 0 30.0000 0
3 8.380075 6.027484 0 13.000000 0 0 30.0000 0
4 8.117634 6.037862 0 13.000000 0 0 30.0000 0
5 8.717578 6.102194 18.242253 13.000000 0 0 0 0
6 8.235365 6.075014 18.118218 13.000000 0 0 0 0
7 8.456258 6.516763 17.025349 13.000000 0 0 0 0
8 8.706969 7.169317 17.016555 13.000000 0 0 0 0
9 8.456812 7.857222 15.514518 13.000000 0 0 0 0
10 8.280751 7.777599 15.664138 13.000000 0 0 0 0
11 8.756094 8.186814 15.359684 13.000000 0 0 0 0
12 8.072241 8.480346 15.529925 13.621897 0 0 0 0
13 8.644599 8.820010 16.043639 15.186864 0 0 0 0
14 8.789204 8.739779 16.077855 15.157092 0 0 0 0
15 8.417893 8.522086 15.861159 15.465897 0 0 0 0
16 8.422415 8.670103 17.284617 15.711638 0 0 0 0
17 7.414206 9.423450 16.388232 15.686373 0 0 0 0
18 7.890915 9.406743 15.451460 16.681479 0 0 0 0
19 7.840446 10.507508 14.476307 16.311760 0 0 0 0
20 7.611747 10.741128 13.594396 17.234454 0 0 0 0
21 7.563631 11.844256 10.007392 18.471548 0 0 0 0
22 7.398136 9.891042 10.000989 19.327728 0 0 0 0
23 5.301084 9.872224 10.082063 20.537381 0 0 0 0
24 5.025130 11.469501 10.002271 22.054193 0 0 0 0

Table 7
Optimal hourly generation scheduling attained with the proposed QOTWFO algorithm for case 1.

Hour (t) Hydropower output (MW) Thermal power output (MW) Power demand (MW)

Ph ð1; tÞ Ph ð2; tÞ Ph ð3; tÞ Ph ð4; tÞ Ps ðtÞ PD (t)

1 86.0001 58.1491 0.0000 200.0937 1025.7571 1370.00


2 87.6260 54.1522 0.0000 187.7553 1060.4666 1390.00
3 76.8456 52.1080 0.0000 173.7333 1057.3131 1360.00
4 74.9050 53.7754 0.0000 156.7917 1004.5279 1290.00
5 77.4350 55.2229 24.9406 178.7421 953.6595 1290.00
6 74.2373 55.4969 25.4511 198.9585 1055.8562 1410.00
7 75.3437 58.3621 29.7766 217.4409 1269.0767 1650.00
8 76.8519 62.6779 29.4882 234.1893 1596.7927 2000.00
9 76.0230 67.1192 33.7931 239.0041 1824.0606 2240.00
10 75.9361 67.2750 33.6769 243.5460 1899.5660 2320.00
11 79.7539 70.1932 34.9670 247.0077 1798.0782 2230.00
12 76.2497 71.6663 35.7098 256.1219 1870.2523 2310.00
13 80.3946 73.1554 36.3664 271.1305 1768.9531 2230.00
14 82.1058 72.8471 37.6188 271.3117 1736.1165 2200.00
15 80.4824 71.8638 39.8981 273.9549 1663.8008 2130.00
16 80.8547 72.3498 36.8572 275.9275 1604.0109 2070.00
17 74.3665 75.0860 40.5894 276.0297 1663.9284 2130.00
18 77.7141 73.0069 44.0942 283.7577 1661.4270 2140.00
19 77.2166 76.0904 46.6918 280.3442 1759.6571 2240.00
20 75.3382 75.3049 49.2278 287.7105 1792.4185 2280.00
21 74.8935 77.6379 50.5911 294.6509 1742.2266 2240.00
22 73.8427 68.9773 52.8206 296.3007 1628.0587 2120.00
23 57.4543 67.5917 54.7160 296.7538 1373.4843 1850.00
24 55.2560 71.7384 56.0643 294.1434 1112.7979 1590.00

16
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

losses are bestowed in Tables 12 and 13 respectively. The power


transmission loss at each interval is estimated according to Eq. (7)
and is included in Table 13. While the optimal hourly hydrother-
mal power generation together with the load demand is shown in
Fig. 16. It is worth noting that the power generation of hydrother-
mal plants are equal to the sum of transmission losses and load
demand at each time horizon. Fig. 17 portrays the obtained hourly
storge volume of each reservoir in which the volume constraints
are met out for all the time intervals. Comparison of statistical re-
sults obtained by different approaches for this test case is given in
Table 14. It seems that the QOTWFO algorithm outclasses the other
compared algorithms. The best optimum fuel cost obtained by us-
ing the QOTWFO algorithm is 41317.0113 $/day which is lesser than
that of others. The optimal solution of QOTWFO satisfies all the
hydraulic and thermal constraints of SHTS problem with the
consideration of VPL and power losses.
Fig. 12. Optimal hourly hydrothermal power generation for case 1.
7.7. Case 4

A large-scale SHTS problem with four hydro and ten thermal


power plants is examined in this case. The number of control var-
iable for this case is 336, which is quite large. This case analyses a
more practical hydrothermal power system by taking into account
the VPL effects and multi-fuel sources. The multi-fuel coefficients
and power bounds of thermal plants for this case are provided in
Ref. [50]. Tables 15 and 16 present the optimal scheduling solution
obtained by QOTWFO algorithm for this case, where a number of
whirlpools of 8 is adequate to deal with SHTS problem solving.
While Figs. 18 and 19 show the obtained hourly hydrothermal po-
wer generation and the hourly reservoir storage volume for the
QOTWFO. All figures demonstrate that the QOTWFO algorithm
provides the feasible optimal planning solution with the satisfac-
tion of all reservoir storage limits and the equality real power
balance constraint. The best fuel cost solution obtained by different
heuristic approaches are itemized in Table 17. As it is evident from
Fig. 13. Hourly reservoir storage volume for case 1. the table that the optimal scheduling solution for this case is ob-
tained in 16.04s, where the minimal thermal power generation cost
is 5521.7312 $. Based on the statistical results compared, it can also
approaches. be remarked that the QOTWFO algorithm with the quasi-
oppositional learning strategy capitulates better quality solutions
7.6. Case 3 and greater stability than the TWFO, GA [50], SPPSO [50], DE [50],
TLBO [50], SOS [50], QRSOS [50], IMO [49] and QRIMO [49]
This case studies a four hydro and three thermal power plants approaches.
with the examination of VPL effects and power transmission losses.
The power loss coefficients for this case are given in the following: 7.8. Case 5

2 3 The case 5, proposed by Ref. [54] is a real and extremely complex


þ0:34 þ0:13 þ0:09 0:01 0:08 0:01 0:02
6 7 power system with 5 thermal plants, 4 hydro plants, 16-bus, and 35
6 7
6 þ0:13 þ0:14 þ0:10 þ0:01 0:05 0:02 0:01 7 transmission lines. This case analyses a more practical hydrother-
6 7
6 7 mal power system by taking into account the VPL effects. Here,
6 þ0:09 þ0:10 þ0:31 þ0:00 0:11 0:07 0:05 7
6 7 NewtoneRaphson AC load flow is employed to determine the po-
6 7
B ¼ 6 0:01 þ0:01 þ0:00 þ0:24 0:08 0:04 0:07 7 wer losses and active power flows in the transmission lines. The
6 7
6 0:08 0:05 0:11 0:08 þ1:92 þ0:27 0:02 7 planning period is 24 h with 4-h time intervals. The voltage is
6 7
6 7 assumed to be 1:05:00 p:u: at the slack bus. The base MVA is taken
6 0:01 0:02 0:07 0:04 þ0:27 þ0:32 þ0:00 7
6 7 as 100. The thermal plants are connected at 1, 4, 5, 8 and 15 buses,
4 5
0:02 0:01 0:05 0:07 0:02 þ0:00 1:35 and the hydro plants are connected at 10, 12, 14 and 16 buses. The
water inflows of 10th and 14th hydro reservoirs are respectively
4
 10 MW 1 assumed as 10  104 m3 =h and 6  104 m3 =h. The transmission line
parameters, generation limits, load demands, and fuel cost co-
B0 ¼ ½  0:75  0:06 0:70  0:33 0:27  0:77  0:01  106 efficients of thermal and hydro plants are taken from Ref. [54]. The
optimal scheduling solution attained by the proposed QOTWFO
approach is provided in Tables 18 and 19. The hydropower gener-
B00 ¼ 0:55MW
ation for each reservoir is estimated on the basis of the hourly water
The optimal hourly water discharges of hydro reservoirs, and the discharge rate, water inflow rate and storge volume at previous
optimal power scheduling of hydrothermal plants and transmission interval using Eq. (5). As it is evident that the best optimal solution
17
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 8
Comparison of solutions attained among different heuristic approaches for case 1.

Algorithm Minimum cost ($/day) Maximum cost ($/day) Mean cost ($/day) Standard deviation Computation time (s)

GA [8] 932734.00 939734.00 936969.00 e e


GWPSO [15] 930622.50 951253.20 940036.30 e 129.10
FEP [10] 930267.92 931396.81 930897.44 e e
CEP [10] 930166.25 930927.01 930373.23 e e
IFEP [10] 930129.82 930881.92 930290.13 e 1033.20
PSO [14] 928778.00 935012.00 933085.00 e e
GCPSO [15] 927288.40 972658.30 936717.10 e 182.4
BCGA [25] 926922.71 929451.09 927815.35 e 64.51
QEA [23] 926538.29 930484.13 928426.95 e 6.32
APSO [34] 926151.54 e e e e
RCGA [25] 925940.03 926538.81 926120.26 e 57.52
PSO [15] 925618.50 e e e e
LWPSO [15] 925383.80 927240.10 926352.80 82.90
SPSO [33] 925308.86 927203.63 926185.32 694.00 45.3
GSO [20] 925173.35 928239.73 927024.68 e e
DE [12] 923991.08 e e e 14.50
RCGA [26] 923966.29 924232.07 924108.73 e e
DE [23] 923234.56 928395.84 925157.28 e 8.69
EPSO [38] 922904.00 924808.00 923527.00 e e
MDE [28] 922555.44 e e e 45
IPSO [30] 922553.49 e e e 38.46
DRQEA [23] 922526.73 925871.51 923419.37 e 7.98
DNLPSO [36] 922498.00 923580.00 922837.00 e 37
CRQEA [24] 922477.14 e e e e
HCRO-DE [41] 922444.79 922936.17 922513.62 e e
MAPSO [34] 922421.66 923508.00 922544.00 e 64
TLBO [18] 922373.39 922873.81 922462.24 e e
RCGA-AFSA [26] 922339.63 922362.53 922346.32 e 11
SPPSO [33] 922336.31 923083.48 922668.45 226.50 16.3
MDNLPSO [36] 922336.30 923404.50 922676.20 e 35
CPSO [38] 922328.64 922564.52 922367.85 e 12.9
CSO [22] 922316.16 922373.10 922326.70 180.37 36
TWFO 922293.99 922763.85 922529.63 141.4294 6.852
QOTWFO 922284.05 922659.69 922498.56 111.8556 5.723

Table 9
Optimal hourly hydro discharge and hourly spillage attained with the proposed QOTWFO algorithm for case 2.

Hour (t) Hydro discharges (104 m3 ) Hydro spillages (104 m3 )

Qh ð1; tÞ Qh ð2; tÞ Qh ð3; tÞ Qh ð4; tÞ Sh ð1; tÞ Sh ð2; tÞ Sh ð3; tÞ Sh ð4; tÞ

1 10.008114 9.108340 0 13.000037 0 0 30.0000 0


2 9.463818 8.629201 0 13.000021 0 0 30.0000 0
3 9.101140 6.109139 0 13.000075 0 0 30.0000 0
4 11.251612 8.168916 0 13.000091 0 0 30.0000 0
5 7.814401 6.534032 13.967971 13.000031 0 0 0 0
6 7.376899 6.052410 0 13.000000 0 0 29.9143 0
7 6.706093 6.161230 13.045165 13.000058 0 0 0 0
8 6.955214 6.096695 14.898000 13.000003 0 0 0 0
9 6.868040 6.449538 13.031363 14.561978 0 0 0 0
10 10.306087 8.379906 13.039907 13.734317 0 0 0 0
11 9.854965 6.452908 14.238183 13.599111 0 0 0 0
12 9.764425 8.838945 16.329962 13.000042 0 0 0 0
13 7.742092 8.387575 15.325198 13.330939 0 0 0 0
14 9.242289 8.552485 16.282014 13.000026 0 0 0 0
15 7.684512 8.694635 18.765449 15.032342 0 0 0 0
16 9.523675 6.473501 15.915034 15.526444 0 0 0 0
17 9.771880 10.401501 15.300153 15.357097 0 0 0 0
18 7.736690 9.325138 15.059336 19.434498 0 0 0 0
19 7.221414 9.542480 15.178981 18.308632 0 0 0 0
20 6.390686 11.306300 15.065217 19.183816 0 0 0 0
21 6.822090 11.491041 10.087698 18.319829 0 0 0 0
22 6.576964 9.929869 10.428897 18.477691 0 0 0 0
23 5.402041 10.855375 11.144455 18.087446 0 0 0 0
24 5.414857 10.058843 10.621730 19.201713 0 0 0 0

18
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 10
Optimal hourly generation scheduling attained with the proposed QOTWFO algorithm for case 2.

Hour (t) Hydropower output (MW) Thermal power output (MW) Power demand (MW)

Ph ð1; tÞ Ph ð2; tÞ Ph ð3; tÞ Ph ð4; tÞ Ps ðtÞ PD (t)

1 86.0344 67.4602 0.0000 200.0940 1016.4115 1370.00


2 83.2496 64.4878 0.0000 187.7554 1054.5073 1390.00
3 80.9916 50.4483 0.0000 173.7337 1054.8263 1360.00
4 88.6414 64.2076 0.0000 156.7921 980.3589 1290.00
5 71.4457 54.7218 40.0066 178.7420 945.0839 1290.00
6 68.5190 51.8922 0.0000 198.9582 1090.6306 1410.00
7 64.3456 52.5751 39.9927 217.4411 1275.6455 1650.00
8 66.7518 52.6181 37.8018 234.1890 1608.6393 2000.00
9 67.0520 55.9446 40.6893 248.1521 1828.1620 2240.00
10 87.0513 68.5586 41.1012 255.3556 1867.9333 2320.00
11 85.7771 57.6484 39.5842 253.5546 1793.4357 2230.00
12 85.4229 72.1184 36.1010 249.1456 1867.2122 2310.00
13 74.7752 69.3451 41.3436 252.2518 1792.2842 2230.00
14 84.5789 70.5424 40.2710 248.9290 1755.6787 2200.00
15 75.7569 71.5172 32.3341 267.7631 1682.6287 2130.00
16 87.0930 58.3332 42.9126 272.8436 1608.8177 2070.00
17 88.1910 79.1393 45.2799 271.3362 1646.0536 2130.00
18 76.1077 71.8987 47.5003 298.7893 1645.7039 2140.00
19 72.4376 71.4317 47.9878 291.9038 1756.2391 2240.00
20 66.0590 77.1018 49.4992 294.3305 1793.0095 2280.00
21 69.4371 76.0776 52.0044 285.6106 1756.8703 2240.00
22 67.7869 68.8650 54.0847 283.0844 1646.1789 2120.00
23 58.3951 70.8849 56.5235 277.4272 1386.7694 1850.00
24 58.8474 65.9667 57.1129 279.9077 1128.1654 1590.00

of the QOTWFO approach with a fuel cost of 147629.4269 $ and a


power loss of 1.7422 p.u. is attained in 72.69 s. Figs. 20 and 21 reveal
the hydrothermal power generations and reservoir storage vol-
umes for entire time span attained from the proposed algorithm.
According to these figures, there are no violations in the storage
volume and real power balance constraints. Table 20 presents the
active power flow on transmission lines, which confirms that the
line flows satisfy their power transfer capability limits.

7.9. Case 6

To further evince the feasibility of proposed methodology, a real


extremely large-scale test system is analysed. This test case con-
siders ten hydro and forty thermal power plants with VPL effects.
All input data is taken from real-world hydrothermal power sys-
tems or broadly utilized in the literature. For instance, the data of
Fig. 14. Optimal hourly hydrothermal power generation for case 2. forty thermal plants is taken from Taipower system [55] comprising
of coal-fired, oil-fired, gas-fired, gas turbine, diesel, combined cycle,
and nuclear units, and the data of hydropower plants is referred
from Ref. [56]. The control variables for this test case are the hourly
discharge rates of ten hydro power plants and the hourly power
generation of forty thermal plants. There are 1200 decision vari-
ables comprising 10  24 water discharges and 40  24 thermal
power generations. Besides, the VPL effects are considered for all
the thermal plants, which increase the complexity of this case.
The optimal hourly water discharges and hydrothermal power
output acquired by the QOTWFO approach are respectively pre-
sented in Tables 21 and 22. Fig. 22 illustrates the optimal hourly
hydro and thermal power generation for this case. It is apparent
that the sum of power generation of hydrothermal plants is equal to
the power demand at each time interval. Fig. 23 displays the hourly
reservoir storage volume obtained by the proposed algorithm,
which confirms that the related constraints of reservoirs are
satisfied.
The results obtained by QOTWFO are compared with TWFO and
Fig. 15. Hourly reservoir storage volume for case 2. mixed-integer linear programming (MILP) [56] approaches as
shown in Table 23. The optimal fuel cost of 2197895.62 $/day is
acquired by the QOTWFO, which is the lowest among the other
19
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 11
Comparison of solutions attained among different heuristic approaches for case 2.

Algorithm Minimum cost ($/day) Maximum cost ($/day) Mean cost ($/day) Standard deviation Computation time (s)

IFEP [10] 933949.25 942593.02 938508.87 e 1450.90


RCGA [26] 930565.242 931427.212 930966.356 e e
DE [23] 928662.84 e e e 8.7
DE [30] 928236.94 e e e e
HDE [29] 927895.81 e e e e
IPSO [30] 925978.84 e e e 31
MDE [28] 925960.56 e e e 27
DRQEA [23] 925485.21 e e e 7.5
MAPSO [34] 924636.88 927,431 926,496 e e
TWFO 924547.3877 925164.6323 924895.3946 192.7358 7.56
QOTWFO 924543.7617 924965.2784 924716.0975 119.4861 6.92

Table 12
Optimal hourly hydro discharge (104 m3 ) attained with the proposed QOTWFO al-
gorithm for case 3.

Hour (t) Qh ð1; tÞ Qh ð2; tÞ Qh ð3; tÞ Qh ð4; tÞ

1 10.180977 6.065997 21.987999 6.498631


2 7.591795 6.918039 24.729237 6.273363
3 8.990401 6.122649 22.966484 6.062728
4 8.489022 6.383847 17.733242 6.022676
5 8.729323 6.534007 17.93368 6.260673
6 8.699258 7.163852 18.117394 8.375143
7 9.414461 7.738752 17.983248 12.968928
8 8.594637 6.419247 15.951429 10.770834
9 7.400663 6.975586 19.060527 12.655844
10 8.452698 8.730136 17.193196 15.96929
11 9.338363 9.073975 16.660958 17.352442
12 8.097146 6.675241 17.540812 16.051379
13 9.0458 9.433206 18.256695 19.129452
14 9.157266 11.2082 18.636458 18.750259
15 8.403724 9.074835 18.133473 17.698622
16 6.719078 9.726442 19.072139 16.739831
Fig. 16. Optimal hourly hydrothermal power generation for case 3.
17 9.745011 13.955476 15.346467 18.98403
18 7.243356 6.068435 18.964287 14.184108
19 5.628209 8.501042 12.982795 18.061096
20 9.979468 10.189268 10.258503 20 approaches. Furthermore, the standard deviation and computation
21 5.166913 6.307979 13.180408 20 time of the QOTWFO algorithm are notably less. Accordingly, the
22 7.010962 9.995276 11.379835 20 comparison between QOTWFO and other approaches authenticates
23 5.956584 12.177756 10.59167 20
that the QOTWFO algorithm predominates the other methods in
24 6.964885 10.560757 18.983806 17.499694
solving the extremely large-scale SHTS problem.

Table 13
Optimal hourly generation scheduling and power loss attained with the proposed QOTWFO algorithm for case 3.

Hour (t) Hydropower output (MW) Thermal power output (MW) Power loss (MW) Power demand (MW)

Ph ð1; tÞ Ph ð1; tÞ Ph ð1; tÞ Ph ð1; tÞ Ps ð1; tÞ Ps ð2; tÞ Ps ð3; tÞ PL (t) PD (t)

1 86.7525 50.5953 30.8167 135.1914 102.6104 124.9078 229.5196 10.3941 750.00


2 72.8333 57.0498 6.4723 128.6463 174.9724 124.9079 229.5194 14.4064 780.00
3 81.0056 53.2153 14.6990 121.6574 174.9346 124.9076 139.7595 10.1813 700.00
4 77.6900 56.4610 39.4357 115.2705 102.6076 124.9076 139.7599 6.1329 650.00
5 78.1286 58.2733 39.1325 133.2954 102.6671 124.9077 139.7593 6.1666 670.00
6 77.3470 62.4321 38.6813 174.9216 102.6229 124.9077 229.5197 10.4332 800.00
7 80.5214 65.1510 39.2018 235.0203 102.6006 209.8154 229.5192 11.8298 950.00
8 76.3915 56.7844 45.4586 217.7648 102.6656 294.7239 229.5196 13.3116 1010.00
9 69.9814 61.1181 35.0402 242.6886 174.8317 294.7234 229.5192 17.8975 1090.00
10 77.3829 72.0287 41.7299 275.6453 102.5998 294.7232 229.5195 13.6294 1080.00
11 83.1478 73.8813 42.5575 287.2409 102.6596 294.7234 229.5197 13.7331 1100.00
12 76.6030 59.8764 39.9075 276.8199 102.5905 294.7231 319.2792 19.7995 1150.00
13 82.7730 75.7228 38.9252 299.4834 102.6822 294.7237 229.5191 13.8333 1110.00
14 84.1853 82.6614 38.1050 295.4474 102.5445 209.8159 229.5192 12.2763 1030.00
15 80.3819 72.5307 40.1664 287.1264 102.6200 209.8156 229.5195 12.1614 1010.00
16 69.1813 74.8771 37.2006 280.6562 174.9197 209.8152 229.5190 16.1703 1060.00
17 88.6584 85.6404 50.1054 296.0034 102.5563 209.8158 229.5198 12.2978 1050.00
18 73.1397 48.7287 39.5332 261.7754 102.5265 294.7234 319.2753 19.6987 1120.00
19 60.3583 63.1126 54.1818 294.2012 174.9557 209.8156 229.5192 16.1483 1070.00
20 89.3769 70.1435 54.9211 305.9466 102.5941 209.8157 229.5195 12.3173 1050.00
21 56.0353 49.7901 57.1163 301.1336 102.5004 124.9076 229.5199 10.9995 910.00
22 71.4474 70.4155 58.0664 300.0447 102.4998 124.9078 139.7599 7.1377 860.00
23 63.3743 76.1128 57.8900 292.4977 102.5123 124.9075 139.7592 7.0504 850.00
24 71.8684 68.1571 48.7369 269.0104 84.4084 124.9078 139.7597 6.2142 800.00

20
7.10. Statistical analysis

The non-parametric statistical tests such as Wilcoxon signed


rank, Friedman and Quade tests [57] are applied to authenticate the
significance of the results obtained from the QOTWFO and TWFO
algorithms for the SHTS problem.
Wilcoxon signed rank test is based on the signed difference
between the performance measure of two algorithms. The hy-
potheses for the nonparametric test consider the median of algo-
rithm based on the performance scores in n independent trials. The
test statistic for the Wilcoxon signed rank test is expressed as
follows:

W ¼ minðW þ ; W  Þ (46)
Fig. 17. Hourly reservoir storage volume for case 3.
If the value of W is lesser than the critical value of the quantiles

Table 14
Comparison of solutions attained among different heuristic approaches for case 3.

Algorithm Minimum cost ($/day) Maximum cost ($/day) Mean cost ($/day) Standard deviation Computation time (s)

EP [9] 45063.00 e e e e
SPSO [33] 44980.32 49166.68 46112.85 e 139.7
DE [28] 44526.11 e e e e
IPSO [17] 44321.24 e e e e
CABC [42] 43362.68 e e e 21
RCGA [26] 42886.35 43261.91 43032.33 e e
ALO [19] 42833.91 e e e 55.63
DE [23] 42801.04 e e e 21
SPPSO [33] 42740.23 44346.97 43622.14 e 32.7
MHDE [29] 42679.87 e e e 40
MDE [28] 42611.14 e e e 125
PSO [16] 42474.00 e e e e
CSA [17] 42440.57 e e e e
TLBO [18] 42385.88 42441.36 42407.23 e e
HDE [28] 42337.30 e e e 48
GSO [43] 42316.39 42379.18 42339.35 e 617.36
QTLBO [45] 42187.49 42202.75 42193.46 e e
QOGSO [43] 42120.02 42145.37 42130.15 e 625.07
MHDE [28] 41856.50 e e e 31
DRQEA [23] 41435.76 e e e 18
TWFO 41318.9838 41386.4218 41358.8003 20.8360 12.84
QOTWFO 41317.0113 41372.9157 41346.1419 15.5570 11.53

Table 15
Optimal hourly hydro discharge and hydro power generation scheduling attained with the proposed QOTWFO algorithm for case 4.

Hour (t) Hydro discharges (104 m3 ) Hydropower output (MW)

Qh ð1; tÞ Qh ð2; tÞ Qh ð3; tÞ Qh ð4; tÞ Ph ð1; tÞ Ph ð2; tÞ Ph ð3; tÞ Ph ð4; tÞ

1 8.063175 7.382745 24.597379 13.126849 76.0840 58.6194 13.3625 201.0741


2 10.150149 7.801363 23.571911 13.140778 86.9654 61.3170 13.9828 188.6412
3 8.415986 6.851039 23.294973 13.143271 77.7750 56.5771 10.7402 174.3960
4 10.176139 6.522951 20.996902 13.134642 85.7394 55.7659 23.9093 157.1227
5 6.530406 6.274147 24.117189 13.106419 64.0834 54.9768 1.8988 171.9982
6 10.191993 9.657064 18.57121 13.160675 84.3011 73.7803 34.0942 185.4518
7 9.254635 7.953593 19.349485 13.197892 79.4559 63.2830 29.4351 197.7719
8 8.197149 7.19359 19.225778 13.130179 74.0020 58.4615 29.6468 206.0984
9 9.840562 8.052413 17.468463 13.603762 82.6379 63.7217 37.4023 221.7335
10 8.274181 7.962516 17.483248 13.775638 75.4779 63.8078 37.2184 228.2935
11 9.409074 7.973088 18.182315 13.168863 82.6110 64.4755 34.7063 228.9299
12 8.078426 8.016292 18.114081 13.539422 75.7356 64.7238 35.0470 237.9497
13 8.702423 9.05041 17.536712 13.766451 80.1650 69.9262 38.6086 243.5998
14 8.867945 8.412076 17.11751 14.016328 82.0173 66.7596 40.6758 249.1764
15 6.786503 7.079045 17.480324 13.52126 69.1077 59.7031 40.4120 248.6835
16 8.442189 9.014568 16.631732 15.805328 80.8877 70.6359 44.0927 271.5114
17 8.317536 8.900387 16.944972 15.884034 80.2438 68.8743 43.3329 273.7095
18 7.796377 9.379915 15.082066 17.022827 76.8422 69.2636 48.3378 282.9563
19 7.795454 9.036625 15.578579 17.715965 76.6830 66.1708 48.3531 287.9075
20 6.328197 9.184926 15.15188 16.671128 65.7468 66.1562 49.9231 279.9528
21 7.378394 11.100311 10.284993 17.982884 73.6413 73.4086 52.7780 288.7581
22 6.994164 9.292043 10.946832 19.780316 71.0130 65.0793 54.9673 295.4978
23 5.583423 9.015454 10.239682 20.327998 60.0011 62.9655 55.3742 293.3854
24 5.425521 10.893438 11.121904 21.573798 58.9439 69.5257 57.7916 292.0992

21
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 16
Optimal hourly thermal power generation scheduling attained with the proposed QOTWFO algorithm for case 4.

Hour (t) Thermal power output (MW) Power demand (MW)

Ps ð1; tÞ Ps ð2; tÞ Ps ð3; tÞ Ps ð4; tÞ Ps ð5; tÞ Ps ð6; tÞ Ps ð7; tÞ Ps ð8; tÞ Ps ð9; tÞ Ps ð10; tÞ PD (t)

1 129.9788 113.3099 195.1875 64.5287 150.6430 64.9815 159.2812 161.6870 246.6555 114.6069 1750.00
2 129.8688 113.3101 179.9194 64.9409 155.8923 64.5016 168.8451 170.3398 244.5241 136.9516 1780.00
3 129.8791 113.5693 199.7713 64.9545 144.9728 64.9998 182.9229 115.7316 225.6714 138.0390 1700.00
4 129.5925 113.7303 183.7826 64.6157 149.1090 64.5074 146.7951 110.3068 223.4828 141.5406 1650.00
5 129.7489 113.3312 195.8020 64.7868 156.7341 64.9400 161.8917 111.2697 231.9071 146.6312 1670.00
6 129.8069 113.5921 201.4491 64.6272 147.0493 64.9449 169.6948 184.6364 231.2380 115.3341 1800.00
7 129.9059 113.7386 177.2851 64.6239 158.0896 119.9922 183.4385 177.4063 248.4570 207.1171 1950.00
8 129.5853 113.9869 194.7319 119.7974 147.1845 119.6423 199.3145 184.6961 241.2900 191.5624 2010.00
9 129.8050 113.1926 203.9927 64.7163 155.5228 119.6337 181.4986 184.3025 264.9720 266.8682 2090.00
10 129.6493 113.5288 198.2914 119.9637 154.8411 119.8446 201.8506 173.1037 245.4438 218.6853 2080.00
11 129.9106 113.9964 180.4080 119.7228 149.1247 119.6815 204.7734 158.2728 237.7411 275.6460 2100.00
12 129.6989 113.4620 176.2352 119.7755 149.7629 119.9045 203.6009 184.9654 263.5329 275.6056 2150.00
13 129.9955 113.5820 204.1421 119.7231 156.0365 119.8203 160.8737 180.1165 235.7467 257.6640 2110.00
14 129.7046 113.5564 184.4588 119.5118 143.7778 119.6868 169.7337 172.9999 231.1321 206.8091 2030.00
15 129.6851 113.4444 204.1565 119.8452 151.2290 119.6315 173.4360 151.4414 260.5693 168.6553 2010.00
16 129.6484 113.6121 187.3731 119.6290 150.0580 119.5876 175.4716 151.0839 243.2729 203.1357 2060.00
17 129.6828 113.0098 198.5208 119.9611 148.2880 119.5627 196.1975 181.1051 248.6791 128.8325 2050.00
18 129.8423 113.0020 199.4355 64.6438 145.5876 119.7073 196.2572 172.7326 259.4514 241.9403 2120.00
19 129.8098 113.3986 182.8935 119.5601 141.8565 119.8033 175.3791 171.6417 245.1073 191.4356 2070.00
20 129.9554 113.8441 177.1031 119.8236 149.2950 119.9168 165.4018 155.3616 256.8790 200.6407 2050.00
21 129.9133 113.0992 184.2356 64.8881 140.1566 64.6633 180.1513 153.7534 259.4039 131.1493 1910.00
22 129.9764 113.4602 193.1555 64.5182 142.8546 64.7034 159.3184 183.4973 241.8614 80.0972 1860.00
23 129.7879 113.0044 195.9295 64.7643 152.9090 64.7348 181.5103 173.5669 225.1960 76.8706 1850.00
24 129.5169 95.9602 185.1836 64.8850 148.9494 64.8131 172.3823 121.2950 252.2463 86.4079 1800.00

Fig. 19. Hourly reservoir storage volume for case 4.

QOTWFO demonstrates a notable improvement over TWFO with


5 % significance level.
The average minimum fuel cost error obtained by QOTWFO and
TWFO algorithms for test cases are listed in Table 25. Table 26
presents the ranks, test statistics and p-values computed through
Fig. 18. Optimal hourly hydrothermal power generation for case 4.
the Friedman and Quade tests. It is evident that the computed value
of Friedman and Quade statistics are higher than the critical value,
of Wilcoxon test statistic for n degrees of freedom, the null hy- and the computed p-values are lesser than the defined value of 5 %
pothesis is dismissed. This reflects that the proposed algorithm level of significance. These verify that the null hypothesis can be
excels the other one, with the correlated p-value. Table 24 presents dismissed, which implies there is a critical distinction between the
the W þ , W  and p-values obtained between QOTWFO and TWFO two algorithms.
algorithms for all the test cases. As the table states that the

22
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 17
Comparison of solutions attained among different heuristic approaches for case 4.

Algorithm Minimum cost ($/day) Maximum cost ($/day) Mean cost ($/day) Standard deviation Computation time (s)

GA [50] 5563.1267 5625.5432 5585.5966 30.5777 112.42


SPPSO [50] 5556.4654 5590.6285 5564.6645 14.8913 76.32
DE [50] 5555.4724 5587.1846 5565.6203 15.0980 85.12
TLBO [50] 5550.8441 5575.7559 5553.8335 8.2623 25.63
SOS [50] 5546.8755 5571.2046 5549.7949 8.0690 17.53
QRSOS [50] 5544.2691 5569.3382 5545.2718 5.0138 11.19
IMO [49] 5541.8860 5556.2618 5542.6047 3.2145 10.28
QRIMO [49] 5539.8923 5543.8021 5540.0877 0.8742 7.20
TWFO 5527.2112 5539.9339 5535.8735 3.7604 16.47
QOTWFO 5521.7312 5528.3922 5524.7431 1.8499 16.04

Table 18
Optimal hydro discharge (104 m3 ) and storage volume (104 m3 ) attained with the proposed QOTWFO algorithm for case 5.

Hour (t) Hydro discharges (104 m3 ) Hydro storage volumes (104 m3 )

Qh ð10; tÞ Qh ð12; tÞ Qh ð14; tÞ Qh ð16; tÞ Vh ð10; tÞ Vh ð12; tÞ Vh ð14; tÞ Vh ð16; tÞ

0 (Vh begin ) e e e e 100.0000 80.0000 170.0000 120.0000


1 6.0592 9.8941 7.2274 12.7517 103.9409 76.1651 168.7726 124.3698
2 6.0858 8.8632 6.1965 11.1557 107.8551 73.3876 168.5760 128.2739
3 6.5443 8.4050 5.7383 10.6034 111.3109 71.5269 168.8377 131.8138
4 6.1422 7.9795 5.3128 10.2163 115.1687 69.6895 169.5249 134.8898
5 7.4658 8.4049 5.7382 10.4490 117.7030 68.7504 169.7866 138.5840
6 (Vh end ) 7.7030 6.4533 5.7866 10.8239 120.0000 70.0000 170.0000 140.0000

Table 19
Optimal generation scheduling (p.u.) and power loss (p.u.) attained with the proposed QOTWFO algorithm for case 5.

Hour (t) Ph ð10; tÞ Ph ð12; tÞ Ph ð14; tÞ Ph ð16; tÞ Ps ð1; tÞ Ps ð4; tÞ Ps ð5; tÞ Ps ð8; tÞ Ps ð15; tÞ

1 0.622369 0.695535 0.484075 2.136250 0.704188 0.459442 0.456924 0.505122 0.898290


2 0.632058 0.625091 0.444343 2.015888 1.651272 0.959102 0.459760 0.505279 1.247207
3 0.674498 0.587483 0.424970 1.990181 2.239939 0.957854 0.458365 0.508525 1.248185
4 0.647418 0.550644 0.406187 1.975362 2.151576 0.958562 0.456832 0.507850 2.095569
5 0.753559 0.567904 0.425806 2.034818 2.548519 0.955283 0.459801 0.506396 1.247915
6 0.773592 0.457888 0.428131 2.089303 0.982840 0.959715 0.455910 0.505946 1.246674
Power loss (p.u.) 1.742194
Fuel cost ($/day) 147629.4269
Computation time (s) 72.6902

Fig. 20. Optimal hourly hydrothermal power generation for case 5.

23
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

If the value of qs is zero or negative, the solution with tuned


parameter set is equivalent or better than the solution with the
perturbed parameters. The average fuel cost improvement index
acquired by the proposed QOTWFO for 50 independent runs is
given in Tables 27e29. According to Tables 27 and 28, the average
fuel cost reduces with increase in population size and whirlpool's
number. Besides, it is noted that if the QOTWFO algorithm is
converged, there is no considerable reduction of fuel cost with
further population size and number of whirlpools increments.
Alternately, the average computation time of the algorithm is
raised. The vaulting probability is an important parameter to bal-
ance the exploration and exploitation trade-offs in QOTWFO algo-
rithm. The higher value of this parameter is facilitated in
exploration, whereas lower value in exploitation. To analyse the
sensitivity of vaulting probability on the solution quality, the value
Fig. 21. Hourly reservoir storage volume for case 5. of Vpmax is varied between 0.3 and 0.6 with a step size of 0.1. It is
quite clear from Table 29 that, a high value of Vpmax (0.6) offers the
minimum average fuel cost for large-scale hydrothermal power
7.11. Sensitivity analysis
system, while a low value of this parameter (0.3) provides lesser
average fuel cost for small-scale system.
To study the performance of the suggested QOTWFO approach,
sensitivity analysis is performed by varying the number of popu-
lation size, number of whirlpools and maximum vaulting proba-
7.12. Comparative study
bility limit for all the test cases. The average fuel cost improvement
index is determined as follows:
7.12.1. Stability
Fp  Ft The dissemination of the minimum fuel costs obtained by
qs ð%Þ ¼  100 (47) QOTWFO algorithm over 50 independent trials for all the test cases
Ft
is depicted in Fig. 24. As it can be obvious that the minimum fuel

Table 20
Active power flow in transmission lines (p.u.).

Line no. Starting bus no. Ending bus no. Hour (t) max
Pline

1 2 3 4 5 6

1 1 2 0.0971 0.1197 0.4415 0.2681 0.5469 0.1729 0.6250


2 1 3 0.3705 0.3946 0.4256 0.3927 0.5714 0.1565 0.6250
3 1 5 0.1372 0.2200 0.2311 0.2741 0.2554 0.1193 0.3125
4 1 6 0.0493 0.3943 0.5716 0.6106 0.6129 0.2774 0.6250
5 1 7 0.0501 0.5226 0.5701 0.6065 0.5618 0.2567 0.6250
6 2 3 0.0022 0.0360 0.0633 0.1822 0.1845 0.0726 0.1875
7 2 5 0.3817 0.7443 0.7452 0.7497 0.5186 0.6045 0.7500
8 3 4 0.3209 0.4294 0.4045 0.4735 0.4097 0.5556 0.6250
9 3 5 0.3064 0.3120 0.2832 0.3160 0.4034 0.3101 0.3125
10 4 5 0.0043 0.0456 0.0240 0.0051 0.0234 0.0625 0.3125
11 4 9 0.1084 0.3118 0.2871 0.2419 0.2979 0.2617 0.3125
12 4 10 0.1491 0.1155 0.1251 0.1877 0.1113 0.1048 0.3125
13 4 11 0.1835 0.1480 0.1664 0.0605 0.3824 0.1001 0.4500
14 5 6 0.2512 0.2938 0.2931 0.2338 0.1396 0.0148 0.6250
15 5 9 0.0822 0.0284 0.1377 0.1760 0.1488 0.1878 0.2250
16 6 7 0.1245 0.0961 0.1171 0.1139 0.0283 0.1242 0.1250
17 6 8 0.3125 0.1069 0.1544 0.2656 0.1486 0.3124 0.3125
18 6 9 0.2125 0.0589 0.0480 0.0261 0.2272 0.2712 0.3125
19 7 8 0.0401 0.1820 0.0775 0.0478 0.0901 0.2662 0.6250
20 7 13 0.0015 0.1644 0.1286 0.0667 0.1363 0.1077 0.1750
21 7 16 0.7355 0.7059 0.6318 0.5321 0.6228 0.6590 0.7500
22 8 9 0.0548 0.1773 0.1319 0.0977 0.0346 0.1746 0.1875
23 8 13 0.0477 0.0391 0.1447 0.0974 0.2331 0.2851 0.3125
24 9 10 0.7491 0.2033 0.4735 0.5237 0.6319 0.5525 0.9375
25 9 12 0.3776 0.1881 0.3178 0.2868 0.4140 0.1446 0.6250
26 10 11 0.2957 0.3616 0.2449 0.3087 0.0542 0.2916 0.6250
27 10 12 0.0199 0.1730 0.0812 0.0027 0.0439 0.0343 0.3125
28 11 12 0.2045 0.0718 0.0078 0.0811 0.0933 0.0357 0.2500
29 11 14 0.5099 0.1622 0.1809 0.6497 0.5701 0.6685 0.7500
30 12 13 0.0478 0.3072 0.1018 0.0534 0.0373 0.0965 0.6250
31 12 15 0.0855 0.3746 0.2413 0.1320 0.0672 0.2154 0.4375
32 13 15 0.2011 0.0926 0.1996 0.2839 0.1511 0.1073 0.3125
33 13 16 0.6593 0.5967 0.5753 0.6486 0.6922 0.7034 0.7500
34 14 15 0.0259 0.3218 0.2441 0.1657 0.1443 0.2404 0.6250
35 15 16 0.7414 0.7133 0.7839 0.7941 0.7197 0.7271 0.9375

24
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 21
Optimal hydro and spillage discharges (104 m3 ) attained with the proposed QOTWFO algorithm for case 6.

Hour (t) Qh ð1; tÞ Qh ð2; tÞ Qh ð3; tÞ Qh ð4; tÞ Qh ð5; tÞ Qh ð6; tÞ Qh ð7; tÞ Qh ð8; tÞ Qh ð9; tÞ Sh ð9; tÞ Qh ð10; tÞ

1 9.9573 7.2323 20.7893 13.4358 9.8257 8.0384 16.0358 14.7633 0.0000 27.5462 14.8841
2 9.7263 7.0603 20.1304 14.8979 10.8217 7.9122 16.0492 13.2129 0.0000 27.7887 13.1160
3 8.9175 6.3414 20.8690 13.3274 7.7127 6.1843 18.8176 14.0549 0.0000 27.6285 13.1395
4 7.8869 6.3491 21.7282 13.0203 7.8780 6.2047 15.5285 13.1834 0.0000 28.1559 14.2149
5 7.5099 7.9894 20.2574 13.1122 8.4021 7.9957 21.0459 14.3162 0.0000 29.3885 14.1516
6 8.3042 6.3043 19.4772 13.1477 8.4557 6.1471 21.5918 13.4831 0.0000 28.1498 13.7024
7 8.9008 6.7738 18.7613 13.3802 8.4694 6.7156 19.7692 15.2182 0.0000 27.6673 14.3370
8 7.3164 8.2199 18.6924 13.0352 7.5929 9.2735 19.6736 14.3634 0.0000 28.3018 14.7617
9 9.4729 7.1178 17.0911 13.2796 7.6000 9.1235 18.2527 14.6217 0.0000 27.8457 13.5057
10 7.6157 9.8546 17.0397 13.0488 11.0272 9.4910 18.5211 14.0614 0.0000 27.5386 14.8697
11 8.6427 7.4163 17.6184 14.5387 8.2186 7.2431 18.5754 13.8095 0.0000 27.1659 14.4356
12 7.7847 7.9302 17.4272 12.3268 7.6915 7.4553 18.5327 12.0638 0.0000 27.7294 24.7816
13 9.4253 8.0136 18.3867 14.1514 8.1862 7.5657 19.6355 14.0808 0.0000 27.1400 26.6365
14 10.8101 7.7398 17.8175 13.7585 8.0194 8.0124 19.0966 13.9992 0.0000 27.7354 26.3474
15 7.5492 7.6958 18.2452 14.1288 8.8562 7.9163 19.3246 15.8751 0.0000 27.6970 27.9334
16 7.2393 7.6107 19.0601 13.9440 7.4876 7.6228 20.5720 15.9409 0.0000 28.0517 27.5344
17 7.8474 9.1637 18.4035 16.1097 7.4164 8.5242 19.7959 13.4017 0.0000 27.5698 26.5806
18 8.2207 8.5720 17.4200 17.1440 8.5396 8.7624 18.6651 15.2989 0.0000 28.2033 28.1556
19 6.5861 9.9881 16.4323 16.2000 7.1282 9.9670 17.8189 17.4958 0.0000 27.6632 29.3348
20 6.8417 10.1292 15.3726 15.8823 7.5276 9.5071 16.6270 13.8824 0.0000 27.5546 26.7543
21 6.7050 11.2576 10.0098 18.2672 6.9824 11.1698 10.0099 16.5416 26.9091 0.0000 30.9248
22 6.7890 11.0020 10.0090 18.1424 7.0527 10.0800 10.0074 18.5701 26.9074 0.0000 28.9552
23 8.4422 9.3498 10.0665 20.0802 7.2760 9.5179 10.0733 19.8653 26.9991 0.0000 32.7439
24 6.5086 12.8883 10.0044 19.4607 6.8323 11.5702 10.0040 18.6255 26.9099 0.0000 31.5204

Table 22
Optimal hourly generation scheduling attained with the proposed QOTWFO algorithm for case 6.

Hour (t) Hydropower output (MW) Thermal power output (MW) Power demand (MW)

Ph ð1; tÞ Ph ð2; tÞ Ph ð3; tÞ Ph ð4; tÞ Ph ð5; tÞ Ph ð6; tÞ Ph ð7; tÞ Ph ð8; tÞ Ph ð9; tÞ Ph ð10; tÞ Ps ðtÞ PD (t)

1 85.8181 57.7585 37.3942 203.4162 85.2474 62.2022 54.5481 212.7414 0.0000 213.5310 6607.3430 7620
2 84.4671 57.2140 36.2617 200.0690 88.8511 61.5102 52.4145 187.1898 0.0000 186.3405 7195.6822 8150
3 79.9931 53.9112 30.0607 173.0123 72.8771 51.9640 43.0233 178.2097 0.0000 172.1599 7934.7887 8790
4 73.6403 55.4221 24.3215 153.3310 73.6630 53.6697 53.4522 153.8794 0.0000 161.3579 8447.2629 9250
5 70.7338 66.0960 31.4496 164.3261 76.0389 65.3581 34.3406 164.6648 0.0000 179.0327 8657.9592 9510
6 75.2345 55.4859 34.4979 173.6551 75.8263 53.7276 30.2133 162.0132 0.0000 193.3233 8676.0229 9530
7 78.2463 58.3017 36.6974 184.9651 75.7322 57.2776 38.2453 179.4056 0.0000 214.2526 8726.8763 9650
8 69.2817 66.7228 36.8097 192.5160 70.8977 71.2322 38.2161 175.0129 0.0000 232.7230 8116.5878 9070
9 82.0178 60.3985 41.9291 202.7167 71.7143 69.7585 42.7698 185.4256 0.0000 237.8216 7595.4480 8590
10 72.4678 75.5072 41.3465 207.8410 89.1609 71.3449 40.6699 190.8457 0.0000 262.7818 7048.0342 8100
11 79.8285 62.7252 39.9635 225.3752 76.7741 59.5426 40.2142 196.1670 0.0000 270.2321 7569.1776 8620
12 75.1023 65.9537 40.3117 212.1861 74.0573 61.2052 41.6740 189.6616 0.0000 340.0662 8129.7819 9230
13 85.2794 66.4490 38.6075 232.0996 77.9871 62.1472 38.0683 212.2766 0.0000 346.4208 8460.6645 9620
14 92.1702 65.4784 40.7471 231.9743 77.8486 65.4425 39.8619 216.7085 0.0000 347.0976 8492.6708 9670
15 74.8814 65.9112 40.0628 238.8205 83.5469 65.4824 38.6967 234.6477 0.0000 349.1768 8508.7734 9700
16 73.1477 65.5861 37.6389 240.5940 75.0768 63.8910 31.9611 238.1007 0.0000 348.8092 8355.1943 9530
17 77.6543 73.4100 39.7357 261.2711 74.8061 68.3857 35.2207 223.8222 0.0000 347.6166 8508.0776 9710
18 80.1169 68.6800 42.8530 269.7709 82.3337 68.0679 39.3883 244.1481 0.0000 349.8507 8354.7907 9600
19 68.3568 74.2233 45.5912 264.7367 72.5858 72.3435 41.7234 262.4477 0.0000 348.9087 8069.0828 9320
20 70.2373 73.5256 48.8552 265.3445 75.2623 69.1311 45.7206 240.5619 0.0000 347.1392 7864.2223 9100
21 69.2200 76.8466 51.5083 283.1789 71.2680 75.0708 51.3976 266.1732 1.1478 346.2725 7527.9164 8820
22 70.0364 74.4497 53.3667 281.5878 71.9388 69.6787 53.3985 280.4793 4.3538 345.3233 7175.3869 8480
23 81.8421 66.0399 54.8985 289.4484 73.8474 65.9662 54.8798 285.9288 5.7829 336.4602 6784.9058 8100
24 68.2447 76.3391 56.0682 281.4083 70.8332 72.1048 56.0675 276.4195 9.0823 333.0577 6430.3748 7730

costs acquired by the QOTWFO algorithm are eminently dissemi- time taken by the QOTWFO algorithm is lesser than the TWFO and
nated surrounding the mean cost even for the complex test cases. other compared heuristic approaches. Furthermore, the execution
The distinction in fuel costs is owing to the fact that the solutions time of the QOTWFO and TWFO algorithms increases linearly
acquired by the algorithm fulfil all constraints at each time horizon. rather than exponentially with increase in size and complexity of
Accordingly, the stability of QOTWFO algorithm is proved to solve the SHTS problems. The convergence rate and computational effi-
large-scale SHTS problems. ciency of the proposed algorithm are enhanced by adopting the
quasi-oppositional learning strategy.
7.12.2. Convergence behaviour
The convergence behaviours of six cases illustrated in Fig. 25 7.12.3. Solution quality
evinces that QOTWFO algorithm converges in a lesser number of A close scrutiny of Tables 8, 11, 14, 17 and 23 reveal that the
iterations as compared with the TWFO algorithm and do not trap solutions (minimum, maximum and average costs) obtained are
into a local optimal solution. When Tables 8, 11, 14, 17 and 23, and better than those of the erstwhile approaches, which demonstrates
Fig. 26 are collectively examined, it is obvious that the average CPU the effectiveness of QOTWFO algorithm in acquiring minimum fuel
25
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Table 26
Results of Friedman and Quade tests.

Test case Ranks Weight Ranks

TWFO QOTWFO TWFO QOTWFO

Case 1 2 1 2 2
Case 2 2 1 2.5 2.5
Case 3 2 1 1 1
Case 4 2 1 0.5 0.5
Case 5 2 1 1.5 1.5
Case 6 2 1 3 3
Sum of ranks 12 6 10.5 10.5
Friedman statistic (Fr ) 5 Quade statistic (Fq ) 20.86
p-value 0.0479 0.0060

Fig. 22. Optimal hourly hydrothermal power generation for case 6.

Fig. 23. Hourly reservoir storage volume for case 6.

Table 23
Comparison of solutions attained among different heuristic approaches for case 6.

Algorithm Minimum cost ($/day) Maximum cost ($/day) Mean cost ($/day) Standard deviation Computation time (s)

MILP [56] 2198623.28 e e e 110.77


TWFO 2198290.3817 2198986.2512 2198690.1835 174.0493 102.10
QOTWFO 2197895.6205 2198018.1084 2198436.0597 148.9033 95.36

Table 24
Results of Wilcoxon signed rank test.
Table 27
Test case Wþ W p-value Influence of population size ðNp Þ on the average fuel cost improvement index.
Case 1 1215 60 < 0:00001
Np Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
Case 2 1270 5 0.0025
Case 3 1192 83 < 0:00001 60 0 0 0.0705 0.1206 0. 257 0.306
Case 4 1230 45 < 0:00001 80 0 0 0 0.0545 0.092 0.0673
Case 5 1184 91 < 0:00001 100 0 0 0 0 0 0
Case 6 1175 100 < 0:00001 120 0 0 0 0 0 0

Table 25
Average Error attained for the test cases. Table 28
Influence of number of whirlpools ðNwh Þ on the average fuel cost improvement
Test case TWFO QOTWFO index.
Case 1 39.202 31.004 Nwh Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
Case 2 53.423 33.12
Case 3 5.7750 4.3121 4 0 0 0.1806 0.2267 0.3704 0.7046
Case 4 1.0423 0.5128 8 0 0 0 0 0 0
Case 5 14.33 12.62 12 0 0 0 0 0 0
Case 6 108.53 94.87 16 0 0 0 0 0 0

26
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

the fuel cost minimization of thermal power plants. The quasi-


Table 29 oppositional based learning strategy is introduced in the popula-
Influence of maximum vaulting probability limit ( Vpmax ) on the average fuel cost
tion initialization and iteration vaulting to enhance the conver-
improvement index.
gence rate and computational efficiency of the algorithm. On the
Vpmax Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 other hand, each equality and inequality constraints are indepen-
0.3 0 0 0 0.3206 0.483 0.642 dently addressed to capture the newly generated whirlpool vector
0.4 0.356 0.465 0.234 0.298 0.274 0.466 within the feasible solution region. Besides, modelling of water
0.5 0.422 0.499 0.266 0.211 0.167 0.198 spillage is addressed in the problem formulation to meet out the
0.6 0.532 0.52 0.360 0 0 0
generation limits of hydro power plants. The suggested approach
has been examined on benchmark functions and six test cases with
different dimensions and complexities to eyeball the computa-
costs for all the test cases of the SHTS problems. Moreover, the tional efficacy of the algorithm. The VPL effects, transmission losses
solutions of QOTWFO algorithm satisfy all the equality and and multi-fuel sources are taken into consideration. Comparison of
inequality constraints of the problem. Accordingly, the employ- results using the proposed QOTWFO, TWFO and other erstwhile
ment of different constraints handling mechanisms in the proposed heuristic approaches in the literature has demonstrated that the
algorithm proffers the global optimal feasible solutions. QOTWFO algorithm is better than the compared approaches with
regard to solution quality and convergence rate, which authenti-
cates superior global search ability of the algorithm. The daily fuel
8. Conclusions
cost savings acquired by the QOTWFO algorithm for six test cases
compared to the TWFO are as follows respectively: 9.94 $, 3.63 $,
In this paper, a newly developed quasi-oppositional turbulent
1.97 $, 5.48 $, 123.07 $ and 394.76 $. Moreover, the results of
water flow-based optimization (QOTWFO) has been presented for
nonparametric tests manifest the robustness of the QOTWFO
optimal scheduling the hourly power output of hydrothermal po-
approach. The superiority of QOTWFO approach is demonstrated
wer plants and its vital objective is to discover feasible solutions for

Fig. 24. Dissemination of cost function over 50 independent trials.

27
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Fig. 25. Convergence curves of TWFO and QOTWFO algorithms for all the test cases.

Fig. 26. Average CPU times of QOTWFO algorithm for different test cases.

28
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

on SHTS problem solving, and divulges that the QOTWFO is an 647e56. https://doi.org/10.1016/j.ijepes.2010.11.016.
[18] Roy PK. Teaching learning based optimization for short-term hydrothermal
effectual optimizer to solve complex large-scale constrained opti-
scheduling problem considering valve point effect and prohibited discharge
mization problems. As further research, the QOTWFO algorithm constraint. Int J Electr Power Energy Syst 2013;53:10e9. https://doi.org/
can be obviously protracted to solve multi-objective economic 10.1016/j.ijepes.2013.03.024.
environmental scheduling of hydrothermal plants incorporating [19] Dubey HM, Pandit M, Panigrahi B. Ant lion optimization for short-term wind
integrated hydrothermal power generation scheduling. Int J Electr Power
renewable energy sources. Energy Syst 2016;83:158e74. https://doi.org/10.1016/j.ijepes.2016.03.057.
[20] Zheng JH, Chen JJ, Wu QH, Jing ZX. Reliability constrained unit commitment
Credit author statement with combined hydro and thermal generation embedded using self-learning
group search optimizer. Energy 2015;81:245e54. https://doi.org/10.1016/
j.energy.2014.12.036.
V.P. Sakthivel: Conceptualization, Methodology, Formal anal- [21] Roy PK, Pradhan M, Paul T. Krill herd algorithm applied to short-term hy-
ysis, Investigation, Writing e review & editing, Software, Supervi- drothermal scheduling problem. Ain Shams Eng J 2018;9(1):31e43. https://
doi.org/10.1016/j.asej.2015.09.003.
sion. K. Thirumal: Writing e original draft, Writing e review & [22] Yin H, Wu F, Meng X, Lin Y, Fan J, Meng A. Crisscross optimization based short-
editing, Visualization, Investigation. P.D. Sathya: Formal analysis, term hydrothermal generation scheduling with cascaded reservoirs. Energy
Writing e review & editing. 2020;203:e117822. https://doi.org/10.1016/j.energy.2020.117822.
[23] Wang Y, Zhou J, Mo L, Zhang R, Zhang Y. Short-term hydrothermal generation
scheduling using differential real-coded quantum-inspired evolutionary al-
Declaration of competing interest gorithm. Energy 2012;44(1):657e71. https://doi.org/10.1016/
j.energy.2012.05.026.
[24] Wang Y, Zhou J, Mo L, Ouyang S, Zhang Y. A clonal real-coded quantum
The authors declare that they have no known competing inspired evolutionary algorithm with Cauchy mutation for short-term hy-
financial interests or personal relationships that could have drothermal generation scheduling. Int J Electr Power Energy Syst 2012;43(1):
appeared to influence the work reported in this paper. 1228e40. https://doi.org/10.1016/j.ijepes.2012.06.036.
[25] Kumar S, Naresh R. Efficient real coded genetic algorithm to solve the non-
convex hydrothermal scheduling problem. Int J Electr Power Energy Syst
References 2007;29(10):738e47. https://doi.org/10.1016/j.ijepes.2007.06.001.
[26] Fang N, Zhou J, Zhang R, Liu Y, Zhang Y. A hybrid of real coded genetic al-
[1] Salam MS, Nor KM, Hamdam AR. Hydrothermal scheduling-based Lagrange gorithm and artificial fish swarm algorithm for short-term optimal hydro-
relaxation approach to hydrothermal coordination. IEEE Trans Power Syst thermal scheduling. Int J Electr Power Energy Syst 2014;62:617e29. https://
1998;13(1):226e35. https://doi: 10.1109/59.651640. doi.org/10.1016/j.ijepes.2014.05.017.
[2] Yang JS, Chen NM. Short term hydrothermal coordination using multi-pass [27] Nazari-Heris M, Mohammadi-Ivatloo B, Haghrah A. Optimal short-term gen-
dynamic programming. IEEE Trans Power Syst 1989;4(3):1050e6. https:// eration scheduling of hydrothermal systems by implementation of real-coded
doi: 10.1109/59.32598. genetic algorithm based on improved Mühlenbein mutation. Energy
[3] Fuentes-Loyola R, Quintana VH. Medium-term hydrothermal coordination by 2017;128:77e85. https://doi.org/10.1016/j.energy.2017.04.007.
semidefinite programming. IEEE Trans Power Syst 2003;18(4):1515e22. [28] Lakshminarasimman L, Subramanian S. Short-term scheduling of hydrother-
https://doi: 10.1109/TPWRS.2003.811006. mal power system with cascaded reservoirs by using modified differential
[4] Franco PEC, Carvalho MF, Soares S. A network flow model for short-term evolution. IEE Proc Generat Transm Distrib 2006;153(6):693e700. https://
hydro-dominated hydrothermal scheduling problems. IEEE Trans Power Syst doi.org/10.1049/ip-gtd:20050407.
1994;9:1016e22. https://doi: 10.1109/59.317642. [29] Lakshminarasimman L, Subramanian S. A modified hybrid differential evolu-
[5] Nilsson O, Sjelvgren D. Mixed-integer programming applied to short-term tion for short-term scheduling of hydrothermal power systems with cascaded
planning of a hydro-thermal system. IEEE Trans Power Syst 1996;11:281e6. reservoirs. Energy Convers Manag 2008;49(10):2513e21. https://doi.org/
https://doi: 10.1109/59.486107. 10.1016/j.enconman.2008.05.021.
[6] Guedes LSM, Maia PDM, Lisboa AC, Vieira DAG, Saldanha RR. A unit [30] Hota PK, Barisal AK, Chakrabarti R. An improved PSO technique for short-term
commitment algorithm and a compact MILP model for short-term hydro- optimal hydrothermal scheduling. Elec Power Syst Res 2009;79(7):1047e53.
power generation scheduling. IEEE Trans Power Syst 2017;32(5):3381e90. https://doi.org/10.1016/j.epsr.2009.01.001.
https://doi: 10.1109/TPWRS.2016.2641390. [31] Sakthivel VP, Suman M, Sathya PD. Combined economic and emission power
[7] Zhang J, Lin S, Liu H, Chen Y, Zhu M, Xu Y. A small-population based parallel dispatch problems through multi-objective squirrel search algorithm. Appl
differential evolution algorithm for short-term hydrothermal scheduling Soft Comput 2021;300(3):e106950. https://doi.org/10.1016/
problem considering power flow constraints. Energy 2017;123:538e54. j.asoc.2020.106950.
https://doi.org/10.1016/j.energy.2017.02.010. [32] Chen G, Gao M, Zhang Z, Li S. Hybridization of chaotic grey wolf optimizer and
[8] Orero SO, Irving MR. A genetic algorithm modelling framework and solution dragonfly algorithm for short-term hydrothermal scheduling. IEEE Access
technique for short term optimal hydrothermal scheduling. IEEE Trans Power 2020;8:142996e3020. https://doi.org/10.1109/ACCESS.2020.3014114.
Syst 1998;13(2):501e18. https://doi: 10.1109/59.667375. [33] Zhang J, Wang J, Yue C. Small population-based particle swarm optimization
[9] Basu M. An interactive fuzzy satisfying method based on evolutionary pro- for short-term hydrothermal scheduling. IEEE Trans Power Syst 2012;27(1):
gramming technique for multiobjective short-term hydrothermal scheduling. 142e52. https://doi.org/10.1109/TPWRS.2011.2165089.
Elec Power Syst Res 2004;69(2e3):277e85. https://doi.org/10.1016/ [34] Amjady N, Soleymanpour HR. Daily hydrothermal generation scheduling by a
j.epsr.2003.10.003. new modified adaptive particle swarm optimization technique. Elec Power
[10] Sinha N, Chakrabarti R, Chattopadhyay P. Fast evolutionary programming Syst Res 2010;80(6):723e32. https://doi.org/10.1016/j.epsr.2009.11.004.
techniques for short-term hydrothermal scheduling. Elec Power Syst Res [35] Naik A, Satapathy SC, Abraham A. Modified social group optimizationda
2003;66:97e103. https://doi: 10.1109/TPWRS.2002.807053. meta-heuristic algorithm to solve short-term hydrothermal scheduling. Appl
[11] Basu M. Economic environmental dispatch of hydrothermal power system. Int Soft Comput 2020;95:e106524. https://doi.org/10.1016/j.asoc.2020.106524.
J Electr Power Energy Syst 2010;32(6):711e20. https://doi.org/10.1016/ [36] Rasoulzadeh-akhijahani A, Mohammadi-ivatloo B. Short-term hydrothermal
j.ijepes.2010.01.005. generation scheduling by a modified dynamic neighborhood learning based
[12] Mandal KK, Chakraborty N. Differential evolution technique-based short-term particle swarm optimization. Int J Electr Power Energy Syst 2015;67:350e67.
economic generation scheduling of hydrothermal systems. Elec Power Syst https://doi.org/10.1016/j.ijepes.2014.12.011.
Res 2008;78(11):1972e9. https://doi.org/10.1016/j.epsr.2008.04.006. [37] Yuan X, Wang L, Yuan Y. Application of enhanced PSO approach to optimal
[13] Amjady N, Sharifzadeh H. Solution of non-convex economic dispatch problem scheduling of hydro system. Energy Convers Manag 2008;49:2966e72.
considering valve point loading effect by a new modified differential evolu- https://doi.org/10.1016/j.enconman.2008.06.017.
tion algorithm. Int J Electr Power Energy Syst 2010;32(8):893e903. https:// [38] Wu Y, Wu Y, Liu X. Couple-based particle swarm optimization for short-term
doi.org/10.1016/j.ijepes.2010.01.023. hydrothermal scheduling. Appl Soft Comput 2019;74:440e50. https://doi.org/
[14] Mandal KK, Basu M, Chakraborty N. Particle swarm optimization technique 10.1016/j.asoc.2018.10.041.
based short-term hydrothermal scheduling. Appl Soft Comput 2008;8(4): [39] Dasgupta K, Roy PK, Mukherjee V. Power flow based hydro-thermal-wind
1392e9. https://doi.org/10.1016/j.asoc.2007.10.006. scheduling of hybrid power system using sine cosine algorithm. Elec Power
[15] Yu B, Yuan X, Wang J. Short-term hydro-thermal scheduling using particle Syst Res 2020;178:26e42. https://doi.org/10.1016/j.epsr.2019.106018.
swarm optimization method. Energy Convers Manag 2007;48:1902. https:// [40] Narang N. Short-term hydrothermal generation scheduling using improved
doi.org/10.1016/j.enconman.2007.01.034. e8. predator influenced civilized swarm optimization technique. Appl Soft Com-
[16] Mandal KK, Chakraborty N. Short-term combined economic emission sched- put 2017;58:207e24. https://doi.org/10.1016/j.asoc.2017.04.065.
uling of hydrothermal systems with cascaded reservoirs using particle swarm [41] Roy PK. Hybrid chemical reaction optimization approach for combined eco-
optimization technique. Appl Soft Comput 2011;11(1):1295e302. https:// nomic emission short-term hydrothermal scheduling. Elec Power Compon
doi.org/10.1016/j.asoc.2010.03.006. Syst 2014;42:1647e60. https://doi.org/10.1080/15325008.2014.927026.
[17] Swain R, Barisal A, Hota P, Chakrabarti R. Short-term hydrothermal scheduling [42] Liao X, Zhou J, Ouyang S, Zhang R, Zhang Y. An adaptive chaotic artificial bee
using clonal selection algorithm. Int J Electr Power Energy Syst 2011;33: colony algorithm for short-term hydrothermal generation scheduling. Int J

29
V.P. Sakthivel, K. Thirumal and P.D. Sathya Energy 251 (2022) 123905

Electr Power Energy Syst 2013;53:34e42. https://doi.org/10.1016/ TPAS.1978.354571.


j.ijepes.2013.04.004. [51] Ghasemi M, Davoudkhani IF, Akbari E, Rahimnejad A, Ghavidel S, Li L. A novel
[43] Basu M. Quasi-oppositional group search optimization for hydrothermal po- and effective optimization algorithm for global optimization and its engi-
wer system. Int J Electr Power Energy Syst 2016;81:324e35. https://doi.org/ neering applications: turbulent Flow of Water-based Optimization (TFWO).
10.1016/j.ijepes.2016.02.051. Eng Appl Artif Intell 2020;92:e103666. https://doi.org/10.1016/
[44] Gouthamkumar N, Sharma V, Naresh R. Disruption based gravitational search j.engappai.2020.103666.
algorithm for short term hydrothermal scheduling. Expert Syst Appl [52] Tizhoosh HR. Opposition-based learning: a new scheme for machine intelli-
2015;42(20):7000e11. https://doi.org/10.1016/j.eswa.2015.05.017. gence. In: International conference on computational intelligence for
[45] Roy PK, Sur A, Pradhan DK. Optimal short-term hydro-thermal scheduling modelling, control and automation and international conference on intelli-
using quasi-oppositional teaching learning based optimization. Eng Appl Artif gent agents, web technologies and internet commerce (CIMCA-IAWTIC'06),
Intell 2013;26:2516e24. https://doi.org/10.1016/j.engappai.2013.08.002. proceeding, vol. 1; 28-30 November 2005. p. 695e701. https://doi.org/
[46] Basu M. Improved differential evolution for short-term hydrothermal sched- 10.1109/CIMCA.2005.1631345.
uling. Int J Electr Power Energy Syst 2014;58:91e100. https://doi.org/ [53] Turgut MS, Turgut OE. Global best-guided oppositional algorithm for solving
10.1016/j.ijepes.2013.12.016. multidimensional optimization problems. Eng Comput 2020;1:1e31. https://
[47] Bhattacharjee K, Bhattacharya A, Dey SHN. Oppositional real coded chemical doi.org/10.1007/s00366-018-0684-5.
reaction based optimization to solve short term hydrothermal scheduling € o
[54] Yaşar C, Ozy €n S. A modified incremental gravitational search algorithm for
problems. Int J Electr Power Energy Syst 2014;63:145e57. https://doi.org/ short-term hydrothermal scheduling with variable head. Eng Appl Artif Intell
10.1016/j.ijepes.2014.05.065. 2020;95:e103845. https://doi.org/10.1016/j.engappai.2020.103845.
[48] Bhattacharjee K, Bhattacharya A, Dey SHN. Real coded chemical reaction [55] Sakthivel VP, Suman M, Sathya PD. Squirrel search algorithm for economic
based optimization for short-term hydrothermal scheduling. Appl Soft Com- dispatch with valve-point effects and multiple fuels. Energy Sour B: Econ Plan
put 2014;24:962e76. https://doi.org/10.1016/j.asoc.2014.08.048. Pol 2020;15(6):351e82. https://doi.org/10.1080/15567249.2020.1803451.
[49] Das S, Bhattacharya A, Chakraborty AK. Quasi-reflected ions motion optimi- [56] Jian J, Pan S, Yang L. Solution for short-term hydrothermal scheduling with a
zation algorithm for short-term hydrothermal scheduling. Neural Comput logarithmic size mixed-integer linear programming formulation. Energy
Appl 2018;29:123e49. https://doi.org/10.1109/TPAS.1978.354571. 2019;171:770e84. https://doi.org/10.1016/j.energy.2019.01.038.
[50] Das S, Bhattacharya A, Chakraborty AK. Solution of short-term hydrothermal [57] Derrac J, Garcia S, Molina D, Herrera F. A practical tutorial on the use of
scheduling problem using quasi-reflected symbiotic organisms search algo- nonparametric statistical tests as a methodology for comparing evolutionary
rithm considering multi-fuel cost characteristics of thermal generator. and swarm intelligence algorithms. Swarm Evol Comput 2011;1:3e18.
Arabian J Sci Eng 2018;43:2931e60. https://doi.org/10.1109/ https://doi.org/10.1016/j.swevo.2011.02.002.

30

You might also like