Received 23 January 2024, accepted 21 February 2024, date of publication 4 March 2024, date of current version 15 March 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3372862

Analysis of Wind Characteristics for Grid-Tied

Wind Turbine Generator Using Incremental
Generative Adversarial Network Model
RAMESH KUMAR BEHARA AND AKSHAY KUMAR SAHA , (Senior Member, IEEE)
Discipline of Electrical, Electronic, and Computer Engineering, University of KwaZulu-Natal, Durban 4041, South Africa
Corresponding author: Akshay Kumar Saha ([email protected])
This work was supported in part by the University of KwaZulu-Natal.

ABSTRACT Wind attribute analysis is a crucial aspect of meteorological and environmental research,
with applications ranging from renewable energy generation to weather forecasting. However, existing
models encounter several challenges in accurately and comprehensively characterizing wind positions. In this
context, the proposed Incremental Tuned Generative Adversarial Network model (incremental GAN model),
based on an unsupervised learning approach, introduces innovative solutions to overcome these challenges
and enhance the precision and reliability of wind position analysis. This research aims to enhance the
reliability and efficiency of wind energy generation by analyzing wind conditions and providing accurate
data for decision-making. It introduces an Incremental GAN that refines parameters based on various factors.
This GAN model learns and predicts these parameters over time, improving its performance. It incorporates
advanced techniques like a 2-level fused discriminator and self-attention for precise predictions of wind
characteristics. The GAN model generates important parameters such as droop gain, which influences
generator output in response to load or generation changes, aiding grid stability. It also optimizes the
frequency control of different types of generators in the presence of wind farms. The model continuously
monitors wind farm conditions, adjusting power injection into the grid as needed for efficient and reliable
wind energy utilization.

INDEX TERMS Renewable energy system (RES), wind energy integration (WEI), power system stability
(PSS), unsupervised learning (USL), incremental tuned generative adversarial network (IGAN), doubly fed
induction generator (DFIG), synchronous generator (SG).

I. INTRODUCTION both national and international initiatives aimed at harness-

Renewable energy sources (RES) are becoming increasingly ing electricity from renewable resources [3], [4]. However,
crucial in meeting the world’s future energy requirements. renewable energy sources possess a distinctive characteristic
They are gradually replacing conventional energy sources, that they are often unpredictable and intermittent. This unpre-
thereby mitigating the risks associated with the widespread dictability can lead to significant challenges concerning the
use of traditional energy sources [1]. With growing con- stability of electricity grids [5], [6]. Wind energy, one of the
cerns about climate change, the imperative to generate clean most abundant renewable energy sources available, is gaining
and eco-friendly power has never been more pressing [2]. substantial traction [7]. While wind energy currently con-
The intensive consumption of electrical energy, coupled with tributes only a fraction of the world’s total power demand,
escalating hydrocarbon prices and a heightened environmen- its growth is accelerating rapidly [2]. Additionally, the cost
tal consciousness, has prompted numerous nations to launch per watt of electricity generated from wind is declining in
tandem with its expansion [8]. This trend underscores the
The associate editor coordinating the review of this manuscript and increasingly prominent role of wind energy integration (WEI)
approving it for publication was Yonghao Gui . in our global transition toward cleaner and more sustainable
2024 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License.
VOLUME 12, 2024 For more information, see https://creativecommons.org/licenses/by/4.0/ 38315
 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

TABLE 1. Nomenclature table.

power generation, and power system stability (PSS) [9], solution for countries across the globe seeking to diversify
[10], [11]. their energy sources. This appeal stems from the inherent
Wind energy is harnessed by employing turbines that are renewable and abundant nature of wind energy [3]. However,
linked to electrical generators. In the quest for optimizing it’s important to note that the conversion of wind energy is
power generation, variable-speed wind turbines employ adap- inherently stochastic due to the unpredictable variations in
tive shafts to track changes in wind speed effectively [8], [12], wind speed [14], [15].
[13]. Additionally, the significance of generators cannot be Doubly Fed Induction Generators (DFIGs) find their niche
overstated, as they play an equally vital role in facilitating the in high-power wind generation systems (WGSs), especially
seamless integration of wind power into the grid. Wind energy in the context of variable-speed wind turbines [16]. In wind
conversion systems (WECSs) have emerged as an intriguing turbine applications where DFIGs are extensively utilized,

38316 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

the rotor’s rotational speed is dynamically adjusted to match Artificial Neural Network (ANN) techniques are gaining
varying wind speeds. DFIGs offer the versatility of operating prominence as an alternative to conventional approaches or
in both hypo-synchronous and hyper-synchronous genera- as integral components of integrated systems. ANN tech-
tor modes, making them a preferred choice for modern niques offer several advantages, including the ability to
grid systems with fluctuating wind speeds. Compared to learn from examples, tolerance for noisy and incomplete
their fixed-speed counterparts like synchronous and induction data, the capacity to handle non-linear problems, and the
generators, DFIGs present several advantages, including cost- capability to perform rapid forecasting and generalization
efficient converters, minimal power losses, the capability for once trained [23]. Artificial intelligence finds applications in
four-quadrant operation, and a wide range of speed operating diverse sectors, including engineering, economics, medicine,
capabilities [17], [18], [19], [20]. These attributes make them and the military. It has proven valuable for the modeling, iden-
particularly well-suited for variable wind speed operations in tification, optimization, forecasting, and control of complex
contemporary grid systems. and nonlinear systems [6]. Table 1 includes an elaborate list
Nonetheless, controlling grid-connected DFIGs presents of nomenclature utilised in the study.
a complex challenge due to the need to regulate active and The main contributions of this research encompass the
reactive powers between the grid and DFIG-based wind creation of an Incremental Tuned GAN capable of predicting
generation systems (WGS) [21]. The essence of employing crucial wind energy parameters, resulting in more precise
variable-speed technology in wind turbines lies in the ability decision-making. This model achieves superior predictive
to harness power efficiently across a wide range of wind accuracy by incorporating advanced techniques like adaptive
speeds, thereby optimizing power generation for varying discriminators and self-attention mechanisms. Additionally,
wind conditions [6], [22], [23]. In the realm of WECSs, the research focuses on optimizing grid stability through
DFIGs, especially when integrated with upward-oriented parameters like droop gain and fine-tuning control strategies
horizontal-axis wind turbines, play a dominant role in con- for various types of generators in the context of renewable
verting mechanical energy into electrical power. The control energy sources like wind farms. Importantly, the research
of DFIGs, achieved through the manipulation of wind turbine establishes a continuous monitoring and adjustment pro-
blade pitch angles, allows for precise regulation of the wind cess, ensuring efficient wind energy utilization by iteratively
turbine generator’s output power, even in the face of higher assessing wind farm conditions and adapting power injection
wind speeds [20], [24]. This capability ensures the efficient into the grid when necessary [21].
and reliable harnessing of wind energy for sustainable power The manuscript is structured into different segments.
generation. Section II provides an explanation of the procedures and the
Recent advancements in wind turbine and generator tech- challenges associated with analyzing wind speed and direc-
nologies have led to a significant reduction in the energy tion. In Section III, an effective method for detecting wind
costs associated with wind farms (WFs) [24]. However, the speed direction is presented, which involves using incremen-
widespread deployment of wind farms can pose challenges tal GAN, the results and findings of the method are illustrated
to power system stability (PSS). Additionally, the variabil- in Section IV, and Section V describes the conclusion and
ity of wind generator output can lead to fluctuations in the future work of the research.
frequency of power systems, impacting the operation of tur-
bines. While wind farms are equipped with frequency relays II. MOTIVATION
to disconnect them in response to frequency disturbances, The motivation behind this research is to harness wind
mass disconnections of generating units can result in power energy effectively and sustainably at various wind behavior
oscillations [25]. as explained in sub-section B. By developing an advanced
To address these challenges and enhance the stability predictive model and optimizing grid stability, the study aims
of smart grid-integrated wind power systems, various arti- to enable the widespread use of clean wind power. This
ficial intelligence (AI) techniques have been extensively research aspires to contribute to a greener and more reliable
applied in power electronic converters and induction motor energy future, reducing our dependence on fossil fuels and
drives. These AI techniques, including fuzzy logic (FL), mitigating environmental impacts.
machine/deep learning, expert systems (ES), neural net-
works (NN), robotics, genetic algorithms (GA), and natural A. LITERATURE REVIEW
language processing, emulate human intellectual functions, The authors aimed to enhance control in DFIG-based wind
enabling computers to achieve self-healing capabilities. energy systems using AI-integrated Fractional Order Robust
Machine and deep learning methodologies, in particular, have Control (FFORC) [3]. The FFORC technique reduces chatter-
the potential to provide affordable, accessible, and highly ing, provides smooth power tracking, and exhibits minimal
accurate diagnostic tools for addressing complex optimiza- steady-state error, surpassing the Sliding Mode Control
tion issues such as variable wind speeds, non-linear load (SMC) system. Potential limitations of FFORC should be
variations, power dispatch, and load control at the point of explored further for comprehensive assessment in practical
coupling [23], [26], [27], [28]. applications.

VOLUME 12, 2024 38317

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

The researchers in [8] aimed to highlight the effective- synchronized operation of the wind farm (WF) and the power
ness of artificial organic controllers (AOCs) in controlling grid with the objective of frequency management, albeit with
variable-speed wind generators, specifically the doubly fed the trade-off of increased complexity that entails the use of
induction generator (DFIG) which results in higher power optimization techniques.
output, reduced pitch angle fluctuations, enhanced power In their study, the authors of reference [32] analyzed
efficiency, smoother startup, benefiting grid stability, blade the impact of frequency response control on the interaction
durability, and sensitivity to wind variations but this model between the wind energy conversion system (WECS) and the
required further exploration of practical issues and scalability grid. They put forth a hybrid active damping technique for the
considerations. wind farm (WF) to alleviate the presence of lightly damped
In [5], the researchers aimed to improve control of Brush- modes that may arise. Nevertheless, this alternative approach
less Doubly-Fed Induction Generators (BDFIGs) in wind may prove to be ineffective in situations where the operational
energy systems by introducing an adaptive fuzzy PID con- point is variable.
troller as compared to traditional control methods which The research scholars in [33] learned that when the grid fre-
enhanced the power generation quality and quantity under quency drops below the rated condition, virtual inertia control
varying conditions, addressing limitations of fixed-parameter and WECS de-loading are needed to increase active power
PID controllers. output. Hence, the researchers discussed advanced method-
In [23], the research authors aimed to investigate ologies like predictive control and artificial intelligence can
sub-synchronous oscillation (SSO) issues in grid-connected improve this combination by determining the back-to-back
wind farms, particularly in wind turbines with doubly fed converter reference quantities.
induction generators (DFIGs). This research offers insights The design aspects and building of a DFIG are described
into abnormal wind turbine operations and potential solutions in detail using AC/DC/AC pulse width modulation (PWM)
through advanced control technologies, but this model leads voltage source electronic regulators in the rotor side circuit
to an increase in the cost and complexity of the generating and vector control used to both the GSC and the RSC without
unit. taking any abnormal condition into account [34], [35], [36],
The authors developed a DFIG-based wind energy sys- [37]. The comparison of proportional integral control and
tem [24] and designed a generator speed controller to improve proportional distribution control for the regulation of reactive
speed regulation in response to varying wind speeds. It com- power in a wind farm is the exclusive focus of [38] and [39],
bines system modeling and fuzzy logic-based PID control, which concludes that proportional integral is more resilient
showing that the fuzzy logic controller outperforms the tradi- than proportional distribution control. While a supervisory
tional PID controller, enhancing generator speed regulation system manages the active and reactive power of the entire
in the WECS but its implementation was limited due to its wind generation system, a machine-learning control scheme
complexity. confirms that set points at the level of the wind turbines are
In [27] the researchers developed a novel method for reached [40].
microgrid frequency control using DFIG and battery energy The limitations of the existing method can be overcome
storage system (BESS), ensuring battery charge and stability using the proposed Incremental Tuned Generative Adver-
after disturbances which preserves battery charge, and sta- sarial Network (GAN) method. The proposed method over-
bilizes microgrid frequency effectively but this model faced comes the limitations caused by characterizing wind posi-
difficulty, that was identified in real-life conditions. tions accurately in some unsupervised learning approaches
The authors in [29] investigated underestimated natural by a wind attribute analysis. The capability to predict the
ventilation through small openings, particularly single and wind characteristics is significant where some methods lag
double-sided ones, aiming to distinguish between effective at this phase which is overcome using the 2-level fused
ventilation/purging flow rate (PFR) and bulk air flow rate discriminator and self-attention. The proposed model filters
(AFR). This model offers insights into natural ventilation, the parameters based on the following factors over time,
highlighting underestimations in low wind pressure scenar- performance enhancement and adaptability. The wind farm
ios and introducing ventilation effectiveness. However, this conditions are continuously monitored by the Incremental
model could affect ventilation rates in real-world scenarios, GAN monitors that permit optimal tuning to power injection
potentially limiting its broader applicability. into the grid and provide an efficient and reliable wind energy
In [30], the authors predicted wind speeds for individ- utilization.
ual turbines in a wind farm using a model that analyzes
both time and spatial characteristics. The model effectively B. CHALLENGES
used spatiotemporal features, achieving accurate wind speed Wind behavior is characterized by its inherent complex-
predictions for future periods, and outperforming existing ity, marked by continual fluctuations in wind speed,
models. However, this model faced high consumption over shifts in wind direction, and the presence of turbulence.
time. Effectively predicting and responding to these dynamic
The authors of reference [31] established that the dis- changes in real-time presents a substantial challenge for AI
tributed model predictive control (DMPC) enables the algorithms [8].

38318 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

FIGURE 3. Air velocity variation in an ideal model of a wind turbine.

FIGURE 1. Architecture of the proposed wind turbine farm model.

decision-making. To achieve this, the authors proposed an

incremental GAN model tailored for wind energy manage-
ment as shown in Fig. 1, that utilizes parameters related to
climate, load, wind speed, and wind penetration to predict
gain parameters. The model undergoes incremental train-
ing and incorporates advanced components like an adaptive
2-level fused discriminator and a self-attention mechanism
to amalgamate data from various sources, resulting in more
accurate predictions of wind characteristics and power gen-
FIGURE 2. Operational concept of the wind energy system based on eration. For the purpose of estimating the wind resource
doubly fed induction generator. potential for the existing wind farm, the researchers in [41]
determined an average wind speed of 6.9 m/s at the wind
-AI-driven systems are tasked with the intricate job of turbine mast site. This allowed the meteodyne software
overseeing DFIG to maximize energy capture, all while to produce trustworthy simulation results. Accordingly, the
upholding grid stability. The pursuit of achieving optimal present authors conducted their research using wind speeds
control and synchronization between the wind turbine and the ranging from 8 to 12 m/s. The design wind speed, Vdesign ,
grid presents a complex and non-trivial challenge [6]. should be 1.4 times the annual mean wind speed, Vamws ,
-DFIG-based wind turbines are prone to faults and failures, in accordance with the IEC61400-2 standard [42]. The
which have the potential to diminish their efficiency and Incremental Tuned GAN additionally generates important
overall lifespan. AI-based systems must possess the ability features, including droop gain and optimized gain parameters.
to detect and diagnose these faults proactively, facilitating The droop gain parameter regulates the speed at which a
timely maintenance interventions [20]. generator adjusts its output voltage and frequency in response
-Expanding AI-based solutions to encompass large wind to changes in load or generation, contributing to grid stability
farms equipped with multiple turbines presents its own set by redistributing the load among multiple WF generators.
of challenges. Coordinating and optimizing the simultaneous The research also focuses on parameter tuning for optimized
operation of multiple turbines within a farm necessitates the control of DFIG and Synchronous Generators (SG), partic-
application of advanced AI algorithms [25]. ularly important for grid stability when renewable energy
Wind turbines function in challenging environmental con- sources like wind farms are present. The model continuously
ditions, enduring extremes such as temperature fluctuations, evaluates wind farm conditions, and if the criteria for power
high humidity levels, and exposure to saltwater in off- generation are not met, it iteratively reenters the Incremental
shore installations. AI systems must demonstrate robustness, Tuned GAN for further refinement, ensuring the efficient
ensuring their resilience and reliability under these harsh utilization of wind energy resources.
circumstances [23].
-With the rising significance of AI in the realm of renew- A. MODELING OF WIND ENERGY SYSTEMS
able energy, it becomes imperative to address ethical concerns Fig. 2 illustrates the operational concept of the wind energy
about AI-driven decision-making and its potential conse- system based on DFIG. The system configuration is depicted
quences for local communities and wildlife [20]. in the block diagram, where the stator of the DFIG is directly
linked to the grid, while its rotor is connected through back-
III. METHODOLOGY to-back converters. Before devising a nonlinear controller for
The primary objective of this research is to assess fluctuations this wind energy system, it is crucial to first comprehend
in wind conditions and enhance the efficiency and relia- and derive the nonlinear mathematical models for both the
bility of wind energy generation and management through Rotor Side Converter (RSC) and the Grid Side Converter
the provision of precise and current wind data for informed (GSC) systems. The subsequent subsections will delve into

VOLUME 12, 2024 38319

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

a comprehensive discussion of the mathematical models for And, therefore, coefficient of power can be computed by [44]:
wind turbines. Power extracted by the rotor
Cp =
Total power of the wind stream
B. WIND TURBINE MODELING
4 ρSV (V1 ) − (V2 ) (V1 + V2 )
1 2 2
At the Gottingen Institute in Germany, in the late 1920s, = (10)
2 ρSV
1 3
Albert Betz developed wind machine theory. A. Betz deter-
mined the maximum power a turbine rotor could generate The interference ratio b is another way to express the ratio of
from wind is 59.26% of the kinetic energy available in V2 to V1 .
the wind. Consider the ideal wind rotor at rest in a fluid
atmosphere, as seen in Fig. 3 [43]. The wind speed through V2
b= (11)
the turbine rotor is defined as V (m/s), where the down- V1
wind cross-sectional area (S2 ) is greater than the upwind So, to express force as a function of interference ratio, we can
cross-sectional area (S1 ). The Betz equation addresses wind say ‘‘2’’ as:
turbine efficiency restrictions by slowing the upstream wind 1  
speed (V1 ) to its downstream speed (V2 ) due to viscous and F = ρSV12 1 − b2 (12)
2
pressure drag on the rotor blades. The rotor extracts mechani-
cal energy by lowering air stream kinetic energy from upwind For any given interference factor b, we can express the
to downwind or by breaking the wind flow. Assuming air flow extractable power P as:
is incompressible, the conservation of mass equation is [43]: 1  
P = ρSV13 1 − b2 (1 + b) (13)
m = ρS1 V1 = ρS2 V2 = constant (1) 4
The power coefficient can be written as follows in terms of b:
Euler’s Theorem calculates wind force on the rotor as:
1 
dV Cp = 1 − b2 (1 + b) (14)
F = ma = m = ρSV (V1 − V2 ) (2) 2
dt
By differentiating ‘‘(14)’’ with respect to ‘b’, we obtain:
Incremental work done by the air stream: dE = Fdx

Rotor power content of the wind stream Pr in watts [43]: d Cp 1 d h  i 1
= 1 − b2 (1 + b) = (1 − 3b) (1 + b)
dE dx db 2 db 2
Pr = =F = FV (3) (15)
dt dt
By, equating (dbp ) =0, we can establish
d C
Substituting the value of F from ‘‘(2)’’ in to ‘‘(3)’’, the power
extracted by rotor Pr :
1
(1 − 3b) (1 + b) = 0
Pr = ρSV 2 (V1 − V2 ) (4) 2
1 V2
Ideal turbine power extraction Pm with V2 = 1 3V1 [43]: (16)

b= =
3 V1
Pm = Pupwind − Pdownwind (5) This suggests that the downwind speed should be one-third of
the upwind speed for the wind turbine to operate at its best.
Wind power as a rate of change of kinetic energy:
Hence, the maximum coefficient of performance [44]:
1 2
2 mV1 − 12 mV22 1   "  2 # 
= ṁ V12 − V22

Pw = 1 1 1
1t 2 Cp_max = 1− 1+
1 2 3 3
= ρSV (V1 − V2 ) (V1 + V2 ) (6)
2 Cp_max = 59.26% (17)
Equating ‘‘(4)’’ and ‘‘(6)’’ implies that vind speed at the Cp_opt is known as the Betz limit, which specifies the greatest
rotor cross section is the average of upper and lower wind amount of wind energy that may be transformed into rota-
speeds [44]: tional energy by the ideal wind turbine.
1 The authors considered the reference value of air density
ρSV 2 (V1 − V2 ) = ρSV (V1 − V2 ) (V1 + V2 )
2 at 15 degrees celsius and sea level as 1.225 kg/m3. This is the
1 International Standard Atmosphere’s (ISA) value. the rotor
V = (V1 + V2 ) (7)
2 swept area is 18.134 m2 and the power coefficient is 0.48 [45].
By substituting tehvalue of V in ‘‘(4)’’ [44]: The square of the rotor diameter and the cube of the wind
speed increase the value of Pr considerably. Above sea level,
1
Pr = ρSV (V1 − V2 )2 (V1 − V2 ) (8) air density depends on pressure and temperature [46].
4
ρo
 
1 −gz
Pr = ρSV (V1 )2 − (V2 )2 (V1 + V2 ) (9) ρZ = exp (18)
4 RT RT

38320 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

FIGURE 5. The representation of a DFIG in a synchronous d - q reference

frame [47].

DFIG-based wind energy systems can be described in the d-q

FIGURE 4. Schematic representation of the wind turbine characteristics.
reference frame using the following equations [48].

where ρZ is the air density as a function of altitude in kg/m3, Vds = Rs Ids + ϕ˙ds − ωs ϕqs (24)
ρ is the standard sea-level atmospheric density in Kg/m3, R Vqs = Rs Iqs + ϕ˙qs + ωs ϕds (25)
is the air-specific gas constant 287.05 J/Kg-K, g is the gravity
Vdr = Rr Idr + ϕ˙dr − ωr ϕqr (26)
constant 9.81 m/s2 , T is the temperature Kelvin, and z is the
altitude above sea level in meters [43]: Vqr = Rr Iqr + ϕ˙qr + ωr ϕdr (27)
The formula for the tip speed ratio λ is expressed as ϕds = Ls Ids + Lm Idr (28)
follows: ϕqs = Ls Iqs + Lm Iqr (29)
ωt r
λ= (19) ϕdr = Lr Idr + Lm Ids (30)
V
ϕqr = Lr Iqr + Lm Iqs (31)
The tip speed ratio λopt plays a pivotal role in determining
the maximum power output of a wind turbine at a given wind The self-inductance coefficient of the stator and rotor wind-
speed. Specifically, when λ = λmax , the power coefficient Cp ings is shown by Ls and Lr , where Rs and Rr are the stator and
attains its maximum value, denoted as CP = CP−max . In this rotor resistances. The mutual coupling coefficient between
study, the power performance of the WECS is characterized the rotor and stator is represented by the symbol Lm . In the
by the relationship between CP and λ. The aerodynamic d-q park reference frame, the voltage and current components
torque Tr generated within the wind turbine with a gear ratio of the stator and rotor side are designated as Vds , Vqs , Ids ,
G, and the angular speed ωt , is expressed as follows, under- Iqs , Vdr , Vqr , Idr and Iqr provides the electromagnetic torque
lining its significance in the system’s overall performance [3]: equation per unit as stated in the d-q park reference [48].
1 The researchers in [49] displayed the machine modelling at
P= ρCp_max SV13 (20) varying wind speed in phase domain model and d-q refer-
2
Tt ωt ωr ence frames. Additionally, six-phase transformer, grid-side
Tr = = (21) converter, and rotor-side converter modelling has shown an
G G
By equating ‘‘(19)’’, ‘‘(20)’’ and ‘‘(21)’’, which represent the improved reliability, performance, and less pulsing torque
reference rotor speed ωr and power as follows [3]: as the key benefits for asymmetrical six phase doubly fed
induction generator (ASPDFIG). The researchers in [50]
λopt Gv provided a hybrid strategy for DFIG-based wind power sys-
ωr−ref = (22)
R tems that improves control efficiency with quantum process
1 and deep reinforcement learning (QPDRL). The proposed
Pgrid−ref = ηρr 2 CP−max V 3 (23)
2 QPDRL avoided the local optimal solutions, detected critical
where the wind turbine efficiency is denoted as η. phases, and updated the DFIG regulating algorithms online.
The system showed an improved performance, by reducing
C. DFIG SYSTEM MODELING the DC link voltage ripples by 62% and speed overshoot by
In this section, the mathematical model delves into the 99%. Based on the research [48], and [49] outcomes, the
meticulous derivation of the robust control system, which authors considered the six phase DFIG control system should
encompasses both RSC and GSC converters. be the future work space.
Fig. 5 shows the DFIG equivalent circuit in the d-q ref-
erence frame. Assuming the DFIG’s electrical modelling D. CONTROL SYSTEM FOR RSC CONVERTERS
within the d-q reference frame is feasible. The voltage, flux, Two control schemes, the speed controller, and the d - q axis
and power of the grid side and rotor side converters for current controllers, are derived in this subsection.

VOLUME 12, 2024 38321

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

1) SPEED CONTROLLER The derivation process begins with the development of the
The phrase rotor angular speed ωr , in terms of generator speed controller, which is rooted in the dynamic model. The
speed r is written like: pulse width modulation controller is used for controlling RSC
and GSC. This dynamic model in ‘‘(33)’’ is then reformulated
ωr = Pr (32) and expressed as follows [3]:
The rotor aerodynamic torque-speed ( Tr - r ) relationship, d Tem
sometimes known as the DFIG mechanical equation, is as r = + d1 (45)
dt J
follows [3]:
In this context, a key element known as the lumped uncer-
d
J r = Tem − Tr − fr r (33) tainty term is being introduced, which denotes as d1 . This
dt uncertainty term is precisely represented by the equation [3]:
where the phrase electromagnetic torque Tem in terms of rotor
d-q axis fluxes ϕds , and ϕqs is written as [3]: Tr fr r
d1 = − − (46)
J J
Lm
ϕqs Idr − ϕds Iqr − Tr − fr r


Tem =P (34) Now, within the context of ‘‘(45)’’, it’s important to note
Ls
that Tem stands for the control input while J represents the
Taking into consideration the relationship below to simplify
moment of inertia. Furthermore, formally define the speed
the above expression:
tracking error e1 in the following manner [3]:
ϕds = ϕs , ϕqs = 0 (35)
e1 = r − r−ref (47)
The expression that results from combining ‘‘(34)’’ and
‘‘(35)’’ is as follows: The time derivative of h1 can be expressed e˙1 = ˙r − r−ref
˙
Lm Vs By substituting ‘‘(45)’’ in e˙1 to obtain a modified expression
Tem = P Iqr (36) as follows [3]:
ωs L s
Tem
The expression for the d/q stator voltage vectors under con- e˙1 = + d1 − r−ref
˙ (48)
stant stator flux and per-phase resistance is as follows [3]: J
To formulate the speed controller, we opt for a fractional-order
Vds = 0, Vqs = Vs = ωs ϕs (37)
sliding surface, expressed as follows [3]:
The d/q voltages in the rotor axis Vdr , Vqr and stator side Z
active-reactive powers Ps , Qs are expressed in terms of rotor −α
S1 = c1 O e1 + c2 e˙1 (49)
current Ir as follows by combining ‘‘(24)’’, ‘‘(25)’’, ‘‘(26)’’,
‘‘(27)’’ and ‘‘(28)’’, ‘‘(29)’’, ‘‘(30)’’, ‘‘(31)’’: In this context, the variables c1 , and c2 stand as design
d constants for the sliding surface. By applying the fractional
Vdr = R∗r Idr + σ Lr Idr − σ Lr sωs Iqr (38) operator Oα to ‘‘(49)’’, derive the following expression [3]:
dt
d Lm Vs
Vqr = Rr Iqr + σ Lr Iqr − σ sωs Idr + s Oα S1 = c1 e1 + c2 Oα−1 e˙1
∗
(39) (50)
dt Ls
Lm Vs Upon combining ‘‘(48)’’ with ‘‘(50)’’, obtain the modified
Ps = − Iqr (40)
Ls relation as follows:
V2 Lm Vs
Qs = s −
 
Idr (41) Tem
ωs Ls Ls Oα S1 = ce1 + c2 Oα−1 + d1 − ˙ r−ref (51)
J
where σ and s are defined as follows:
Lm ωs − ωr From ‘‘(51)’’, we derive the speed controller as follows [3]:
σ =1− , and s = (42)
Lr Ls ωs

T = Tem−eq  + Tem−s
 em

 
Here, s stands for the DFIG slip. The nonlinear system model 1−α c1

Tem−eq = J r−ref − O
˙

e1 (52)
is described as follows using the equations mentioned above:  c2 
k

r1

Ẋ = A (x) + B (u) = a (x) + 1a + bu + 1bu  Tem−s = J − O1−α sgn (S1 )

(43) 
c2
Ẋ = A (x) + B (u) = a (x) + bu + d (t) (44)
T T ‘‘Equation (52) introduces kr1 as the discontinuous gain
In this case, u = [u1 , u2 ] = Vdr , Vqr , and X = Idr , Iqr ,
 
parameter of the controller. Notably, within the term, Tem−s
a and b represent the best approximation (nominal model) there is
a fractional derivative that considers the magnitude
of A and B respectively; 1a, and 1b denote plant  and input
T O1−α . This specific inclusion plays a pivotal role in damp-
vector uncertainties, whereas d= 1a + 1bu = dId , dIq is ening the chattering phenomena, ultimately contributing to
the system models unknown lumped uncertainty. improved overall control performance.

38322 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

2) D-Q AXIS CURRENT CONTROL E. FORMULATION OF DC LINK MODELING AND

First, the expressions of active power P and reactive power Q GRID-SIDE CONTROL
are expressed as follows in order to derive the d and q axis Throughout the control operation, the DC link voltage is
current controllers: maintained by the DC link capacitor that is linked between
the RSC and GSC converters. Without changing the DC link
ωs Ls
Iqr−ref = Tem−ref (53) voltage, the GSC converter side controls the active power
PLm Vs (P) and reactive power (Q). This can be accomplished by
V2 Lm Vs applying a traditional vector control method by lining up
Qref = s − Idr−ref (54)
ωs Ls Ls the reference frame with the grid or stator voltage vector.
Consequently, Vd = Vs ,and Vq = 0.
With Qref = 0, reference d-axis current is expressed as
As a result, the following formulas are used for P and Q.
Vs  
3
3
Idr−ref = Tem−ref (55) Ps = Vds Ids + Vqs Iqs = (Vds Ids ) (62)
Lm ωs 2 2
 
The following defines the current tracking faults on the d and 3 3
Vds Ids − Vds Iqs = − (Vds Ids )


Qs = (63)
q axes: 2 2
 T  T The ratio of current component Id to active power and current
e2 = eId eIq = Idr − Idr−ref Iqr − Iqr−ref (56)
component Iq to reactive power flowing between the GSC
ė2 = ẋ − ẋref = a (x) + bu + d − ẋref (57) converter and the grid is represented by ‘‘(62)’’, and ‘‘(63)’’.
The fractional order sliding surface S, with c3 ; c4 ; c5 and c6 3
EIos = Vd Id (64)
are the design parameters: 2
where, E represents the DC link voltage, Ios is the stator
c O−α e + c4 R ėId
   R 
S
S = 2 = 3 −α Id (58) side, and Ior is the rotor side DC link current. Additionally,
S3 c5 O eIq + c6 ėIq
the following shift in rotor side DC power equals the GSC
By applying the fractional operator Oa to ‘‘(58)’’, the converter-grid active power.
resultant expression: dE
C = Ios − Ior (65)
dt
c3 eId + c4 Oα−1 ėId
 α   
α O S2
O S= = (59)
Oα S 3 c5 eIq + c6 Oα−1 ėIq Upon combining ‘‘(64)’’ with ‘‘(65)’’, obtain the modified
relation as follows:
Hence, the equivalent control terms 1

3

Ė = Vd Id − Ior (66)
 
c3 O1−α eId
 C 2E
b −1 −a (x) + İ −
  d−ref c4 1
u
ueq = eq1 =  Ė = g (x) Id − Ior (67)
 

C
 1−α
ueq2 c5 O eIq 
b−1 −a (x) + İqr−ref −

c6 1
Ė = (1g (x) + go (x)) Id − Ior (68)
C
(60) 1
 h
−k r2 O1−α sgn(S2)
i Ė = go (x) Id − Ior + dE (69)
b−1 C
 
us1 c4
us = = (61)
−k r3 O1−α sgn(S3 ) Here, dE = 1g (x) Id , represents an uncertainty term.
h i
us2 b−1 c6
To maintain a constant DC link voltage, the mathematical
The sliding condition is achieved by assigning values to model calculates the error by taking the difference between
Setting kr = kr2 kr3 > dmax where dmax denotes the the reference DC voltage and the actual DC link voltage,


maximum limit of the disturbance and it is assumed that this which is expressed as follows [3]:
limit is known. ∗
eE = E − E (70)
Certainly, Fig. 4 provides a schematic representation of
the wind turbine characteristics under specific conditions. The time derivative of the error in the equation above is
At a wind speed of 5 meters per second (5m/s), the wind expressed as follows:
turbine exhibits its operating parameters. The maximum rev- ∗
olutions per minute (RPM) for this wind turbine model is ėE = Ė − Ė (71)
limited to 1500, and it achieves a maximum power output
Upon substituting the expression for Ė into ‘‘(71)’’, the
of 0.5 per unit under these conditions. This figure serves as
resulting expanded expression is as follows [3]:
an essential visual reference to understand the wind turbine’s
behavior and performance at the specified wind speed and its 1 ∗
ėE = go (x) Id − Ior + dE − Ė (72)
corresponding power generation capabilities. C

VOLUME 12, 2024 38323

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

To regulate the DC link voltage effectively, here opt for a frac- 3) WIND SPEED
tional calculus-based sliding surface described by ‘‘(73)’’: Wind speed is a crucial factor in wind energy generation.
Z Higher wind speeds generally lead to greater energy output,
SE = c7 O−α eE + c8 ėE (73) while low or variable wind speeds can reduce power pro-
duction. Wind speed data is essential for optimizing turbine
Upon applying the operator Oα to both sides of ‘‘(73)’’, the operation and ensuring grid integration.
resulting relation is as follows [3]:
4) WIND PENETRATION
Oα S E = c7 eE + c8 Oα−1 ėE (74) Wind penetration refers to the percentage of total electricity
generation supplied by wind energy within a given area or
By combining ‘‘(73)’’ with ‘‘(74)’’, obtain the following grid. High wind penetration can introduce challenges related
expanded relation: to grid stability, voltage regulation, and the need for energy
  storage or backup power sources during periods of low wind
α α−1 1 ∗
O S E = c7 eE + c8 O ėE go (x) Id − Ior + dE − Ė generation.
C
(75) 5) ROTOR SPEED
The rotor speed of wind turbines must be controlled to maxi-
Utilizing ‘‘(75)’’, the control law can be defined as mize energy capture and prevent damage due to overspeed.
follows [3]: Adjusting rotor speed based on wind conditions is essen-
 tial for optimizing power generation and ensuring turbine
Id = Ideq + Ids
longevity.

 
1 1 c7

 ∗
Ė + Ior − O1−α (eE )

Ideq =
go (x) c7 c8 
(76)
6) POWER FACTOR (PF)
1 k

r4

1−α
(sgn (SE )) The power factor represents the ratio of real power (active

 Ids =
 − O
go (x) c8 power) to apparent power in an electrical system. Wind
‘‘In ‘‘(76)’’ c7 , and c8 serves as the design parameters turbines and wind farms often need to maintain a specific
while, kr4 representing the discontinuous switching gain. power factor to ensure grid compatibility and meet grid codes.
This formulation captures the key components of the ‘‘(76)’’, Control systems may adjust the power factor to maintain grid
allowing for precise control system design and parameter stability.
tuning.
7) DFIG ACTIVE AND REACTIVE POWER
F. PRELIMINARIES The active and reactive power output of the DFIG is essential
Climatic conditions in wind farms, along with other input for grid compatibility. Control of these parameters ensures
parameters like load (active and reactive power), wind speed, that wind farms can meet grid requirements and contribute to
wind penetration, rotor speed, power factor (PF) and DFIG grid stability.
active and reactive collectively influence the performance and Overall, these input parameters, when properly monitored
operation of wind energy systems. Here’s how each of these and controlled, allow wind farms to operate efficiently, con-
parameters can impact wind farms: tribute to grid stability, and integrate renewable energy into
the electrical grid effectively. Managing these parameters is
critical for the successful deployment and operation of wind
1) CLIMATIC CONDITIONS
energy systems.
Climatic conditions, including temperature, humidity, air
density, and extreme weather events, directly affect wind
G. DROOP CONTROL MATHEMATICAL MODEL
turbine performance and efficiency. For example, higher
The droop gain parameter regulates the speed at which a gen-
temperatures reduce air density, which can affect energy
erator adjusts its output voltage and frequency in response to
generation. Severe weather events such as storms or ice accu-
changes in load or generation, contributing to grid stability by
mulation can lead to downtime and maintenance needs.
redistributing the load among multiple wind farm generators.
Each DFIG-WT can be depicted as a voltage source con-
2) LOAD (ACTIVE AND REACTIVE POWER)
nected to a shared bus via line impedance denoted as Z . The
Load refers to the amount of electrical power consumed by power transfer from the wind unit to the Point of Common
connected devices or systems. In a wind farm, variations in Coupling (PCC) bus can be expressed as follows [51]:
active power demand can impact the stability of the electrical
UA e−jδp − UB
 
grid. Reactive power is also important for maintaining voltage SA = PA + jQA = U ejδp (77)
levels and ensuring the grid’s reliability. Wind turbines may A Ze−jγ
need to provide both active and reactive power support to U2 UA U B j(γ +δp )
meet load requirements and grid regulations. SA = A ejγ − e (78)
Z Z
38324 VOLUME 12, 2024
R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

In this context, SA and SB correspond to the magnitudes of the

wind unit and PCC bus voltages, while Z and γ respectively
signify the magnitude and phase angle of the line impedance.
Furthermore, Zejγ = R + jX , and δp which characterizes the
power angle. To establish the voltage difference, the model
employs 1U = UA − UB [51].
1
1U = (RPA + XQA ) (79)
UA
1 FIGURE 6. The architecture of the incremental GAN model.
δp = (XPA − RQA ) (80)
UA UB
‘‘Equations (79) and ‘‘(80)’’ illustrate the decoupling of focus on the discriminator’s novelty, adaptive 2-level fusion,
(P − Q) from each other solely through δp − 1U, but this and the integration of the self-attention mechanism. In this
scenario holds only in an ideal case specifically, in cases research, the incremental GAN comprises two main com-
where there is either pure inductance (X = 0) or pure resis- ponents: the generator and the discriminator. The generator
tance (R = 0). When apply (R = 0) to ‘‘(79)’’ and ‘‘(80)’’, aims to create accurate and up-to-date wind data, including
the following outcome emerges [51]: parameters like wind speed, direction, robustness, and power
generation. The discriminator plays a critical role in assessing
PA = UA UB δp X

(81) the quality and authenticity of the generated data, in this
QA = UA 1U X

(82) research, enhancing the discriminator makes it more effective
at distinguishing between real and generated data.
Under these conditions, the conventional droop-sharing
scheme can be deployed to control and regulate the active 1) WORKING OF THE ADAPTIVE 2-LEVEL FUSION
and reactive power output of the wind turbine generator unit DISCRIMINATOR
as follows [51]:
-Level 1 (Primary Discriminator - Discriminator-1):
f = fo − mP (83) Discriminator-1 assesses the predictions made by the gen-
U = Uo − nQ (84) erator (GAN’s Generator). It takes the generated data as
input and compares it to real data, evaluating how closely the
Here fo and Uo parameters are associated with the fre- generated data matches the actual observations. It calculates
quency and amplitude of the wind turbine generator unit a loss based on the discrepancy between real and generated
voltage at no load (set points). The droop coefficients for data. This loss is used as feedback to update and improve the
these parameters can be described as follows: m is assigned generator’s parameters.
to the frequency droop coefficient and n is indicative of the -Level 2 (Secondary Discriminator - Discriminator-2):
voltage droop coefficient [51]. Discriminator-2 operates at a higher level and evaluates the
assessments made by Discriminator-1. It examines not only
m = 1f Pn

(85) the differences between real and generated data but also con-
n = 1U Qn

(86) siders other features, such as the distribution of data points.
It calculates a secondary loss based on the discrepancies iden-
In this context, Pn represents the nominal active power, and
tified by Discriminator-1 and other relevant information. The
Qn represents the nominal reactive power of the wind unit.
outputs (losses) of Discriminator-1 and Discriminator-2 are
Usually, there exist permissible deviations of up to 2% for
fused or combined to create a unified assessment of the incre-
1f and 5% for 1U. It is a standard practice to maintain 1f ,
mental GAN model’s performance. Both Discriminator-1 and
and 1U as equal values across all units.
Discriminator-2 are trained to become more proficient at
1f = mPn1 = m2 Pn2 = . . . . . . . . . = mj Pnj (87) assessing the quality of generated data. The incremental GAN
1U = n1 Qn1 = n2 Qn2 = . . . . . . . . . = nj Qnj (88) model’s generator uses a combination of D1 loss and D2 loss
as feedback to update its parameters and improve its data
Consequently, the coefficients exhibit an inverse relation- generation.
ship with nominal powers, choosing the droop coefficients
following ‘‘(87)’’ and ‘‘(88)’’ guarantees that each unit con- 2) SELF ATTENTION MECHANISM
tributes to power sharing in alignment with its designated The self-attention mechanism is a neural network component
nominal rating [51]. that plays a critical role in enhancing the discriminator’s and
generator’s capabilities by selectively focusing on important
H. ADAPTIVE SELF ATTENTION-BASED GAN NETWORK features in the input data.
As shown in Fig. 6, the authors provide a more detailed It helps improve the model’s understanding of relationships
explanation of how the GAN is employed in the research between different data points, leading to more accurate pre-
to enhance wind energy generation and management, with a dictions. Self-attention mechanisms analyze the relationships

VOLUME 12, 2024 38325

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

between different elements in the data (e.g., features like relatively small inverter, which only channels the air-gap
wind speed, load, and temperature). The self-attention mech- power, denoted as Pδ , into the grid. Additionally, the grid
anisms assign different attention weights to these elements, choke can be sized to accommodate the lower rotor current.
allowing the model to focus more on relevant information However, this doubly-fed induction generator concept does
while downplaying less important details. By attending to the introduce certain unique features, such as direct harmonic
most informative features, the self-attention mechanism helps emissions from the stator, when compared to other types of
the discriminator make better assessments and the generator generators. Nonetheless, the DFIG is a well-established prin-
creates more realistic data. Finally, this study focused on ciple, and its fundamental equations have been extensively
the originality of the discriminator, using the self-attention documented and published in [52], [53].
mechanism, and implementing an adaptive 2-level fusion d9si
mechanism to improve the performance of the incremental usi = rs .isi + , where i = 1, 2, 3 . . . . (89)
dt
GAN model in wind turbine applications. These components d9ri
work together to improve the incremental GAN model’s abil- uri = rr .iri + (90)
dt
ity to generate accurate and realistic data, which is crucial 3
for optimizing wind turbine operations, including parameters P X d9i
Tel = ii . (91)
like droop gain. The generator and the discriminator are 2 dv
i=1
classified as: 60. (fs + fr )
Generator = [Sequence Input Layer (7), Fully Connected n= (92)
P
Layer (64), Relu Layer (1), Fully Connected Layer (1), Tanh
Layer (1)]. (P, is the total count of pole pairs)
Discriminator = [Sequence Input Layer (1), Fully d9d
ud = rs .id + − e 9q (93)
Connected Layer (64), Relu Layer (1), Fully Connected dt
Layer (1), Regression Layer (1)]. d9q
uq = rs .iq + − e 9d (94)
Using a fully connected neural layers with high-dimensional dt
input data might result in a substantial rise in the parameter e = P.mech (95)
count, rendering it computationally expensive and perhaps
impracticable for training. Therefore, the authors have chosen In the (d − q) reference frame, the calculation of active power
to utilise 64 neurons in the fully connected layers to meet (P) and reactive power (Q) is as follows [52]:
the special criteria of the neural network architecture in this P = 3. ud .id + uq .iq


(96)
situation, as it is compatible with high-dimensional data.
Q = 3. ud .iq + uq .id


(97)
IV. CONTROL STRATEGIES FOR GRID-TIED DOUBLY FED The relationship between stator power (Ps ) and rotor power
INDUCTION GENERATOR (DFIG) AND SYNCHRONOUS (Pr ) is expressed in terms of parameters such as the number
GENERATOR (SG) of turns w and winding factor ε [52]:
Incremental tuning of a GAN optimizes the grid-integrated Ps ws εs fs
DFIG and SG systems by training the generator to create = (98)
Pr wr εr fr
accurate data resembling real parameters, while the discrim-
inator assesses and minimizes discrepancies between the Under identical rotor and stator current conditions, the DFIG
generated and real data. Iterative training refines the gen- can provide twice the power output when compared to
erator’s output, improving authenticity. This enhanced data an asynchronous generator equipped with a squirrel-cage
benefits control algorithms, predictive maintenance, and opti- rotor [52].
mization of DFIG and SG systems, enhancing wind energy
P = 2.U1 .I1 . cos φ1 (99)
power generation efficiency and reliability. DFIG and SG are
two types of generators commonly used in wind farms. Harmonics are generated both by the grid-side inverter and
A precise control strategy is essential to ensure these gen- the rotor-side inverter. These switching distortions are subse-
erators operate effectively and contribute to grid stability. For quently transferred to the stator side and directly injected into
DFIG, control strategies may involve adjusting rotor speeds the grid. The fundamental frequency on the grid side remains
and power factors to match grid requirements. SGs, on the constant at 50 Hz, while the harmonics depend on both the
other hand, may use governor control and excitation control pulse number (for current harmonics) and the pulse frequency
to regulate output power and maintain grid frequency. (for voltage harmonics). The rotor harmonics operate under
the same principles, but they transform the grid side. This
A. DFIG CONTROL SYSTEM transformation introduces a phase shift, which is contingent
The DFIG is distinguished by its sturdy and widely rec- on the instantaneous speed and the machine’s inductance.
ognized design. In this configuration, the stator is directly The outcome is the emergence of transformed rotor harmonic
connected to the grid, while the rotor is linked to the mains frequencies, denoted as ftr−rotor which manifests as interhar-
via an inverter. One notable advantage is the use of a monics on the grid side. These inter-harmonics can be derived

38326 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

from the rotor frequency frotor and the mechanical frequency TABLE 2. DFIG specifications/parameters with units.
fmech [52].
ftr−rotor n = frotor n + fmech (100)
(n, which is a multiple of the rotor-side fundamental).
On the grid side, anticipate the presence of harmonics
with sidebands originating from the grid-side inverter, as well
as interharmonics resulting from the transformation process
occurring within the generator itself, courtesy of the rotor-
side inverter. Sidebands of voltage harmonics are generated
due to amplitude modulation effects. Furthermore, inter-
harmonics arise because of load-dependent variations in the
switching of both the grid-side and rotor-side inverters.
V. RESULTS AND DISCUSSIONS
Consequently, during dynamic WEC operations, it’s not
always feasible to establish a direct correlation between rotor The incremental GAN model is used to analyze the variations
switching and grid-side inter-harmonics. Table 2 summarizes in wind conditions, and its effectiveness is compared to that
the DFIG specifications used in the simulation exercise. of alternative approaches.

B. SG CONTROL SYSTEM A. SIMULATION SETUP

In contrast to the induction machine, the SG operates in a The incremental GAN model has been deployed and
fully decoupled manner from the grid using an inverter. The tested using MATLAB, specifically in the 2022b version,
fundamental equations governing the behavior of the SG are on a Windows 10 platform equipped with 8GB of RAM.
as follows [52]: The primary objective behind this setup is to thoroughly
assess the model’s performance and gauge its progress and
d9si
usi = rs .isi + (101) improvements in functionality and results with 10 m/s.
dt
The stator system, as per ‘‘(37)’’ converting into a self-
B. PERFORMANCE ANALYSIS BASED ON TIME
contained, revolving frame of reference leads to:
The authors presented a comprehensive overview of the per-
d9k formance metrics for the incremental GAN models applied
uk = rs .ik + + j.e 9k (102)
dt to the analysis of variations in wind conditions from Fig. 7 to
With Fig. 14.
Figure 7: The incremental GAN approach consistently
e = pmech (103)
demonstrates remarkable performance, achieving frequency
In the (d − q) reference frame, derive the Ud , and Uq voltages values of 50.075Hz, 50.118 Hz, 50.201 Hz, 50.272 Hz,
to calculate the Tt torque [52]. and 50.285 Hz across epochs 10, 20, 30, 40, and 50,
d9d all within a time frame of 0.06∗ 11 = 0.66 seconds. The
ud = rs .id + − e 9q (104) Epoch 10/20/30/40/50 series of events pertains to, a total =
dt
c9q 60 seconds, and time interval = 60 / 1000 = 0.06 seconds
uq = rs .iq + − e 9d (105) (or 60 milliseconds), 0.06∗ 2 = 0.12. This calculation cor-
dt
3 rectly converts the time from seconds to milliseconds. The
Tt = p. 9d .Iq − 9d .Id


(106) training stages of a machine learning model, showcasing its
2
performance at different training iterations.
Harmonics originate exclusively from the grid-side
On the other hand, variations in droop values are associ-
inverter. Their frequencies are contingent upon the switching
ated with control systems, representing changes in voltage or
frequency (about voltage harmonics) and the inverter’s pulse
frequency levels to maintain stability in electrical systems.
number (associated with current harmonics). The magnitude
Figure 8: In this subfigure, the incremental GAN strategy
is dictated by the pulse duty factor, DP [52].
showcases its ability to maintain steady main load (kW) val-
non
DP = (107) ues, recording 53.28 kW, 53.40 kW, 53.41 kW, 53.42 kW, and
non + noff 53.63 kW while adhering to the 0.12 seconds time frame. The
The effective impedance, which includes both the inverter and main load refers to the total electrical demand from various
grid impedance, determines the presence of interharmonics. devices and systems in a power network. It encompasses
Interharmonics are generated solely when there are variations residential, commercial, or industrial loads and represents
in the pulse duty factor due to fluctuations in the load. This overall power consumption. It is distinct from the ‘‘DFIG
occurrence is routine in the operation of WECs due to their rating,’’ which specifically relates to the capacity of a DFIG
variable-speed characteristics. Table 3 summarizes the SG used in wind turbines. The DFIG rating is a characteristic of
specifications used in the simulation studies. the generator, while the main load is the electrical demand

VOLUME 12, 2024 38327

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

TABLE 3. SG specifications/parameters with units.

or consumption that the generator is designed to serve. The

DFIG must be appropriately sized to meet the requirements
of the main load.
Figure 9: The incremental GAN strategy further demon-
strates its effectiveness by consistently producing secondary FIGURE 7. The incremental GAN approach, achieving frequency (Hz)
load (kW) values of 156.65 kW, 157.01 kW, 157.02 kW, response.
157.04 kW, and 157.68 kW throughout the analysis, main-
taining the time frame at 0.12 seconds. Secondary load likely
refers to the electrical load demand that is not considered the
primary load in the system. The mentioned values 156.65 kW
and 157.68 kW for the secondary load, are likely measure-
ments of power consumption. These values represent the
electrical load in kW at various points in the analysis. It’s
important to note that secondary load is not typically used to
describe a system rating. Instead, it refers to the actual power
demand or consumption in the system at specific times during
the analysis.
Figure 10: This subfigure provides insight into the speed
performance of the incremental GAN approach. It achieves
impressive speed values of 1263.63 rpm, 1263.68 rpm,
1264.26 rpm, 1264.36 rpm, and 1263.45 rpm across vari-
ous epochs, all while maintaining a consistent 0.12-second
interval. The x-difference of the graph indicates the time in FIGURE 8. The incremental GAN approach, achieving main load (kW)
response.
seconds and the y-difference indicates the speed in rpm.
Figure 11: Continuing the analysis, the incremental GAN
strategy reveals its capabilities in managing sync condenser maintained consistently throughout the 0.12-second analysis
values. These values consistently show results of 27.33 kvar, period. The x-difference of the graph indicates the time in sec-
30.03 kvar, 30.16 kvar, 30.33 kvar, and 45.71 kvar within the onds and the y-difference indicates the wind turbine voltage
same 0.12-second time frame. The x-difference of the graph in kvar.
indicates the time in seconds and the y-difference indicates Figure 14: The torque produced by the incremental GAN
the sync condenser in kvar. approach appears to be very close, with slight variations
Figure 12: Fig. 12 provides insights into the power per- across different epochs ranging from 10-50. The values
formance of the incremental GAN approach, which is the obtained at different epochs of 10-50 are 98.02 (N-m),
DFIG power. It achieves active power levels of 206.75 kW, 98.97 (N-m), 100.16 (N-m), 100.45 (N-m), and 101.12 (N-m)
207.87 kW, 207.94 kW, 208.07 kW, and 209.57 kW respectively. The x-difference of the graph indicates the
across different epochs, all while maintaining the consistent time in seconds and the y-difference indicates the torque
0.12-second analysis period. The x-difference of the graph in N-m. These results collectively illustrate the robustness
indicates the time in seconds and the y-difference indicates and stability of the incremental GAN models in han-
the wind turbine power in kW. dling various performance metrics across different epochs
Figure 13: The provided reactive power values for the and time intervals during the analysis of wind condition
wind turbine DFIG component in the incremental GAN variations.
approach appear to be very close, with slight variations These results collectively illustrate the robustness and sta-
across different epochs. All three components achieve bility of the incremental GAN models in handling various
voltage levels of 231.63 kvar, 231.66 kvar, 231.67 kvar, performance metrics across different epochs and time inter-
231.78 kvar, 231.806 kvar. These voltage levels are vals during the analysis of wind condition variations.

38328 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

FIGURE 9. The incremental GAN approach, achieving secondary load (kW) FIGURE 11. The incremental GAN approach, achieving sync condenser
response. (kvar) response.

FIGURE 10. The incremental GAN approach, achieving speed (rpm) FIGURE 12. The incremental GAN approach, wind turbine active power
response. (kW) response.

C. SIMULATION RESULTS COMPARISON OF DIVERSE over which the incremental GAN model’s performance is
TECHNIQUES INCLUDING ANN, FOPI, FUZZY, CNN, AND evaluated in the context of wind condition variations.
THE PROPOSED IGAN Impressively, the incremental GAN model outshines the
A rigorous comparative analysis was meticulously carried out standard GAN model by a remarkable margin of 0.00027%,
to gauge the effectiveness of the incremental GAN model attaining a frequency of 50.0747 Hz. This notable improve-
when applied to the intricate task of analyzing wind condition ment underscores the model’s superior performance in accu-
variations. This evaluative process encompassed an exhaus- rately predicting and stabilizing frequency.
tive exploration of an array of diverse techniques and models, Figure 16 (Main Load): Moving to Fig. 16, examine the
including ANN [54], fractional-order PI controllers [54], steady main load in the context of wind condition analy-
fuzzy-based robust control systems [55], deep CNN [55], sis, with the incremental GAN model maintaining the same
and GAN [56]. Each of these approaches was thoughtfully 0.12 second time interval. Demonstrating its superiority over
introduced and meticulously studied within the confines of the conventional GAN model, the incremental GAN model
this research. achieves an outstanding steady main load of 53.63 kW, sur-
passing its counterpart by a significant margin of 0.04%.
1) COMPARATIVE ANALYSIS BASED ON TIME This achievement underscores the model’s ability to
Figure 15 (Frequency): Fig. 15 provides a detailed insight maintain a consistent and efficient main load.
into the frequency achieved by the incremental GAN model Figure 17 (Secondary Load): Fig. 17 focuses on the
during the analysis of wind condition variations while con- secondary load achieved by the incremental GAN model
sistently maintaining a time interval of 0.12 seconds. The during wind condition analysis, with the time interval held at
time interval of 0.12 seconds likely represents the duration 0.12 seconds. Notably, the incremental GAN model surpasses

VOLUME 12, 2024 38329

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

FIGURE 13. The incremental GAN approach, wind turbine reactive power FIGURE 16. Comparative main load (kW) analysis based on Time.
(kvar) response.

FIGURE 17. Comparative secondary load (kW) analysis based on Time.

FIGURE 14. The incremental GAN approach, wind turbine Torque (N-m)
response.
Figure 18 (Speed): In Fig. 18, the authors delve into the
speed achieved by the incremental GAN model for the anal-
ysis of wind condition variations, maintaining the consistent
0.12-second time frame. Impressively, the incremental GAN
model outperforms the GAN model by a significant margin of
0.0689%, reaching an impressive speed of 1261.03 rpm. This
achievement reflects the model’s superior ability to respond
to changes in wind conditions.
Figure 19 (Sync Condenser): Fig. 19 illustrates the sync
condenser achieved by the incremental GAN model dur-
ing the analysis of wind conditions while adhering to a
0.12-second time frame. Remarkably, the incremental GAN
model outperforms the GAN model by an impressive margin
of 6.86%, achieving a sync condenser of 45.71 kvar. This sub-
stantial improvement underscores the model’s effectiveness
in managing synchronization condenser values.
FIGURE 15. Comparative frequency (Hz) analysis based on Time.
Figure 20 (DFIG active power): Fig. 20, explores the wind
turbine power achieved by the incremental GAN model for
the secondary load of the GAN model by an impressive the analysis of wind condition variations, maintaining the
0.04%, achieving a secondary load of 157.68 kW. This result same 0.12-second time interval. Impressively, the incremen-
highlights the model’s capacity to handle secondary load tal GAN model outperforms the GAN model by a significant
fluctuations effectively. margin of 0.3183%, reaching a power output of 212.90 kW.

38330 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

FIGURE 18. Comparative speed (p.u.) analysis based on Time. FIGURE 20. Comparative DFIG Active Power (kW) analysis based on Time.

FIGURE 19. Comparative synchronous condenser (kvar) analysis based FIGURE 21. Comparative DFIG Reactive Power (kvar) analysis based on
on Time. Time.

This result highlights the model’s enhanced efficiency in

harnessing wind energy and converting it into usable power.
The x-difference of the graph represents the time in seconds
and y-difference represents power in kW.
Figure 21 (DFIG Reactive power): Fig. 21, explores the
wind turbine reactive power achieved by the incremental
GAN model for the analysis of wind condition variations,
maintaining the same 0.12-second time interval. Impres-
sively, the incremental GAN model outperforms the GAN
model by a significant margin of 0.06892%, reaching a
power output of 271 kvar. This result highlights the model’s
enhanced efficiency in harnessing wind energy and con-
verting it into usable power. The x-difference of the graph
indicates the time in seconds and the y-difference indicates
the wind turbine power in kvar. FIGURE 22. Comparative DFIG voltage (p.u.) analysis based on Time.
Figure 22 (DFIG voltage): Fig. 22, explores the DFIG sys-
tem voltage achieved by the incremental GAN model for the The x-difference of the graph indicates the time in seconds
analysis of wind condition variations, maintaining the same and the y-difference indicates the DFIG voltage in p.u.
0.12 second time interval. Impressively, the incremental GAN Figure 23 (SG voltage): Fig. 23, explores the SG voltage
model outperforms the GAN model by a significant margin achieved by the incremental GAN model for the analysis of
of 0.0018%, reaching a DFIG system voltage of 1.00829 p.u. wind condition variations, maintaining the same 0.12-second
This result highlights the model’s enhanced efficiency in time interval. Impressively, the incremental GAN model
harnessing wind energy and converting it into usable power. outperforms the GAN model by a significant margin of

VOLUME 12, 2024 38331

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

TABLE 4. A comparative discussion in analyzing wind speed direction

involves evaluating different methods and techniques used to assess
wind direction data.

FIGURE 23. Comparative SG voltage (p.u.) analysis based on Time.

main load and an increase in the secondary load can benefit

the system, it can indeed be advantageous in certain situa-
tions. By reducing the main load to a range of 0-50 kW and
simultaneously increasing the secondary load to a range of
0-450 kW, a more balanced load distribution is achieved. This
balance can lead to improved overall system performance.
With a reduced main load, there’s a lower risk of overload-
ing primary components, and the increased secondary load
helps utilize available capacity efficiently. This approach can
enhance system stability, minimize energy waste, and lead
FIGURE 24. Comparative System voltage (p.u.) analysis based on the
Time. to better resource utilization, contributing to enhanced sys-
tem efficiency and reliability. However, the specific benefits
0.002 reaching an SG voltage of 1.0084 p.u. The x-difference depend on the system’s characteristics and requirements.
of the graph indicates the time in seconds and the y-difference
indicates the SG voltage in p.u. VI. CONCLUSION
Figure 24 (system voltage): Fig. 24, explores the sys- In conclusion, this research represents a significant step for-
tem voltage achieved by the incremental GAN model for ward in the field of wind energy generation, its integration
the analysis of voltage variations, maintaining the same into the power grid, and power system stability manage-
0.12 second time interval. Impressively, the incremental GAN ment. Deploying the Incremental Tuned GAN and harnessing
model outperforms the GAN model by a significant mar- parameters including climate, load, wind speed, and wind
gin of 0.002 reaching a system voltage of 0.0064 p.u. The penetration, have paved the way for enhanced efficiency and
x-difference of the graph indicates the time in seconds and reliability in the utilization of wind energy resources. The
the y-difference indicates the system voltage in p.u. incorporation of adaptive discriminators, self-attention mech-
anisms, and the generation of key parameters like droop gain
2) COMPARATIVE DISCUSSIONS and optimized gain parameters has substantially improved the
Table 4 summarizes a comparative discussion in analyzing accuracy of wind data predictions. Furthermore, our focus
wind speed direction involves evaluating different meth- on grid stability, particularly in the context of DFIG and
ods and techniques used to assess wind direction data. SG, underscores the importance of the seamless integration
It involves an in-depth exploration of each approach’s of renewable energy sources into existing power systems.
strengths, weaknesses, performance, and applicability in the This research’s iterative approach, where the Incremental
context of analyzing wind condition variations. Ultimately, Tuned GAN continuously assesses and refines wind farm
this enhanced performance could lead to increased energy conditions, ensures that power generation remains respon-
production, reduced operational costs, or more reliable power sive to changing circumstances. At 0.12 seconds of analysis
generation, making the proposed model a preferred choice for time, the incremental GAN model achieves specific oper-
wind turbine analysis applications. Here a reduction in the ational parameters, a frequency of 50.0747 Hz, a primary

38332 VOLUME 12, 2024

R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

load demand of 53.38 kW, a secondary load demand of [8] P. Ponce, J. A. Rosales, A. Molina, H. Ponce, and B. MacCleery,
157.68 kW, a generator speed of 1.0110 per unit, a syn- ‘‘Designing a robust controller using SMC and fuzzy artificial organic
networks for brushed DC motors,’’ Energies, vol. 13, no. 12, p. 3091,
chronous condenser rating of 45.71 kvar, and a wind turbine Jun. 2020, doi: 10.3390/en13123091.
power output of 212.90 kW. In the future, the research will [9] C. Mukonza and G. Nhamo, ‘‘Wind energy in South Africa: A review of
be extended to use a six-phase DFIG system to analyse the policies, institutions and programmes,’’ J. Energy Southern Afr., vol. 29,
no. 2, pp. 21–28, Jun. 2018.
system performance. [10] A. R. Nejad, ‘‘Wind turbine drivetrains: State-of-the-art technologies and
future development trends,’’ Wind Energy Sci., vol. 7, no. 1, pp. 387–411,
A. CONTRIBUTION Feb. 2022, doi: 10.5194/wes-7-387-2022.
• Incremental Tuned GAN: This research aims at the [11] WIND FORCE 12 A Blueprint to Achieve 12% of the World’s Electricity
From Wind Power by 2020, Wind Force, European Wind Energy Associa-
development and application of an incremental tuned tion and Greenpeace, Colombo, Sri Lanka, Jun. 2005, pp. 1–52.
GAN technique tailored for wind energy manage- [12] J. F. Manwell, J. G. McGowan, and A. L. Rogers, Wind Energy Explained
ment. This incremental GAN methodology, powered by Theory, Design, and Applications. U.K.: Wiley, 2009. [Online]. Available:
climate, load, wind speed, and wind penetration parame- https://ee.tlu.edu.vn/Portals/0/2018/NLG/Sach_Tieng_Anh.pdf
[13] A. Hemami, Wind Turbine Technology. Boston, MA, USA: Cengage
ters, significantly enhances predictive accuracy through Learning, 2012, doi: 12345XX111009.
continuous training. It employs advanced discrimina- [14] M. Juraeva, J.-H. Lee, and D.-J. Song, ‘‘A computational analysis of the
tors and self-attention mechanisms to refine predictions train-wind to identify the best position for the air-curtain installation,’’
J. Wind Eng. Ind. Aerodynamics, vol. 99, no. 5, pp. 554–559, May 2011,
and generates critical parameters for grid stability and doi: 10.1016/j.jweia.2011.02.004.
optimized wind energy generation. In essence, the Incre- [15] H. Jing, H. Liao, C. Ma, Q. Tao, and J. Jiang, ‘‘Field measurement study
mental Tuned GAN serves as the research’s core tool for of wind characteristics at different measuring positions in a mountainous
valley,’’ Experim. Thermal Fluid Sci., vol. 112, Apr. 2020, Art. no. 109991,
achieving accurate, real-time wind data predictions, thus doi: 10.1016/j.expthermflusci.2019.109991.
improving wind energy management. [16] R. Akbari, A. Izadian, and R. S. Weissbach, ‘‘Quasi self-excited DFIG-
• Control strategies of Grid-Tied DFIG and SG: The pri- based wind energy conversion system,’’ IEEE Trans. Ind. Appl., vol. 57,
mary contribution of grid-integrated wind farms based no. 3, pp. 2816–2824, May 2021, doi: 10.1109/TIA.2021.3067621.
[17] I. Angulo, D. de la Vega, I. Cascón, J. Cañizo, Y. Wu, D. Guerra, and
on the DFIG and an independent SG in this research P. Angueira, ‘‘Impact analysis of wind farms on telecommunication ser-
is to optimize their control strategies for better grid vices,’’ Renew. Sustain. Energy Rev., vol. 32, pp. 84–99, Apr. 2014, doi:
stability and reliability, particularly when incorporat- 10.1016/j.rser.2013.12.055.
[18] N. Naeeni and M. Yaghoubi, ‘‘Analysis of wind flow around a parabolic
ing wind energy. This ensures a smoother transition to collector (1) fluid flow,’’ Renew. Energy, vol. 32, no. 11, pp. 1898–1916,
cleaner energy sources and more efficient grid man- Sep. 2007, doi: 10.1016/j.renene.2006.10.004.
agement, reducing disruptions caused by variable wind [19] X. Chen and J. Z. Xu, ‘‘Structural failure analysis of wind tur-
bines impacted by super typhoon usagi,’’ Eng. Failure Anal., vol. 60,
power. In essence, DFIGs and SGs’ operational charac- pp. 391–404, Feb. 2016, doi: 10.1016/j.engfailanal.2015.11.028.
teristic improvements enhance the integration of wind [20] B. C. Babu and K. B. Mohanty, ‘‘Doubly-fed induction generator for
energy into the power grid, promoting sustainability and variable speed wind energy conversion systems-modeling & simula-
tion,’’ Int. J. Comput. Electr. Eng., vol. 2, pp. 141–147, Jun. 2010, doi:
reliability. 10.7763/ijcee.2010.v2.127.
[21] X. Zhao, Z. Lin, B. Fu, and S. Gong, ‘‘Research on frequency control
REFERENCES method for micro-grid with a hybrid approach of FFR-OPPT and pitch
angle of wind turbine,’’ Int. J. Electr. Power Energy Syst., vol. 127,
[1] International Energy Agency. (2022). International Energy Agency (IEA) May 2021, Art. no. 106670, doi: 10.1016/j.ijepes.2020.106670.
World Energy Outlook 2022. [Online]. Available: https://Www.Iea. [22] C. Wei, M. Han, S. Lin, and X. Li, ‘‘Low voltage ride-through control
Org/Reports/World-Energy-Outlook-2022/Executive-Summary strategy of a large capacity DFIG with grid-side thyristor controlled voltage
[2] D. Byles and S. Mohagheghi, ‘‘Sustainable power grid expansion: regulator,’’ in Proc. IET Conf. Renew. Power Gener. (RPG), Sep. 2011,
Life cycle assessment, modeling approaches, challenges, and oppor- pp. 1–6, doi: 10.1049/cp.2011.0215.
tunities,’’ Sustainability, vol. 15, no. 11, p. 8788, May 2023, doi: [23] B. Lu, S. Li, H. S. Das, Y. Gao, J. Wang, and M. Baggu, ‘‘Dynamic
10.3390/su15118788. P-Q capability and abnormal operation analysis of a wind turbine
[3] N. Ullah, I. Sami, Md. S. Chowdhury, K. Techato, and H. I. Alkhammash, with doubly fed induction generator,’’ IEEE J. Emerg. Sel. Top-
‘‘Artificial intelligence integrated fractional order control of ics Power Electron., vol. 10, no. 4, pp. 4854–4864, Aug. 2022, doi:
doubly fed induction generator-based wind energy system,’’ IEEE 10.1109/JESTPE.2021.3133527.
Access, vol. 9, pp. 5734–5748, 2021, doi: 10.1109/ACCESS.2020.3 [24] E. Haile, G. Worku, A. M. Beyene, and M. Tuka, ‘‘Modeling of doubly
048420. fed induction generator based wind energy conversion system and speed
[4] M. Debouza and A. Al-Durra, ‘‘Grid ancillary services from controller,’’ J. Energy Syst., vol. 5, no. 1, pp. 46–59, Mar. 2021, doi:
doubly fed induction generator-based wind energy conversion 10.30521/jes.854669.
system: A review,’’ IEEE Access, vol. 7, pp. 7067–7081, 2019, doi: [25] J. D. A. de Oliveira, F. K. D. A. Lima, F. L. Tofoli, and C. G. C. Branco,
10.1109/ACCESS.2018.2890168. ‘‘Synchronverter-based frequency control technique applied in wind
[5] Z. Tir, O. P. Malik, and M. N. Hashemnia, ‘‘Intelligent control of a energy conversion systems based on the doubly-fed induction genera-
brushless doubly-fed induction generator,’’ Int. J. Syst. Assurance Eng. tor,’’ Electric Power Syst. Res., vol. 214, Jan. 2023, Art. no. 108820, doi:
Manage., vol. 10, no. 3, pp. 326–338, Jun. 2019, doi: 10.1007/s13198-019- 10.1016/j.epsr.2022.108820.
00772-2. [26] R. K. Behara and A. K. Saha, ‘‘Artificial intelligence control system
[6] I. Yaichi, A. Semmah, and P. Wira, ‘‘Control of doubly fed induc- applied in smart grid integrated doubly fed induction generator-based wind
tion generator with maximum power point tracking for variable speed turbine: A review,’’ Energies, vol. 15, no. 17, p. 6488, Sep. 2022, doi:
wind energy conversion systems,’’ Periodica Polytechnica Electr. Eng. 10.3390/en15176488.
Comput. Sci., vol. 64, no. 1, pp. 87–96, Nov. 2019, doi: 10.3311/ppee. [27] L. A. G. Gomez, A. P. Grilo, M. B. C. Salles, and A. J. Sguarezi Filho,
14166. ‘‘Combined control of DFIG-based wind turbine and battery energy stor-
[7] India Energy Outlook. (2021). India Energy Outlook 2021—Analysis IEA. age system for frequency response in microgrids,’’ Energies, vol. 13, no. 4,
India Energy Outlook. [Online]. Available: www.iea.org/weo p. 894, Feb. 2020, doi: 10.3390/en13040894.

VOLUME 12, 2024 38333

 R. K. Behara, A. K. Saha: Analysis of Wind Characteristics for Grid-Tied Wind Turbine Generator

[28] T.-H. Chien, Y.-C. Huang, and Y.-Y. Hsu, ‘‘Neural network-based supple- [48] M. Badreldien, R. Usama, A. El-wakeel, and A. Y. Abdelaziz,
mentary frequency controller for a DFIG wind farm,’’ Energies, vol. 13, ‘‘Modeling, analysis and control of doubly fed induction generators for
no. 20, p. 5320, Oct. 2020, doi: 10.3390/en13205320. wind turbines,’’ in Proc. ICEENG Conf. ICEENG, 2014, pp. 1–18, doi:
[29] Z. Xiaoxun, X. Zixu, W. Yu, G. Xiaoxia, H. Xinyu, L. Hongkun, 10.21608/iceeng.2014.30383.
L. Ruizhang, C. Yao, and L. Huaxin, ‘‘Research on wind speed behav- [49] N. K. Mishra, Z. Husain, and A. Iqbal, ‘‘Modeling and analysis of novel
ior prediction method based on multi-feature and multi-scale inte- six-phase DFIG through asymmetrical winding structure for disperse gen-
grated learning,’’ Energy, vol. 263, Jan. 2023, Art. no. 125593, doi: eration,’’ Int. Trans. Electr. Energy Syst., vol. 30, no. 12, Dec. 2020,
10.1016/j.energy.2022.125593. Art. no. e12649, doi: 10.1002/2050-7038.12649.
[50] C. Chellaswamy, T. S. Geetha, P. Thiruvalar Selvan, and A. Arunkumar,
[30] T. Kobayashi, M. Sandberg, T. Fujita, E. Lim, and N. Umemiya,
‘‘6-phase DFIG for wind energy conversion system: A hybrid approach,’’
‘‘Numerical analysis of wind-induced natural ventilation for an isolated
Sustain. Energy Technol. Assessments, vol. 53, Oct. 2022, Art. no. 102497,
cubic room with two openings under small mean wind pressure dif-
doi: 10.1016/j.seta.2022.102497.
ference,’’ Building Environ., vol. 226, Dec. 2022, Art. no. 109694, doi:
10.1016/j.buildenv.2022.109694. [51] M. S. Marhaba, S. Farhangi, H. Iman-Eini, and R. Iravani, ‘‘Reactive
power sharing improvement of droop-controlled DFIG wind turbines in
[31] Y. Guo, H. Gao, Q. Wu, J. Østergaard, D. Yu, and M. Shahidehpour, a microgrid,’’ IET Gener., Transmiss. Distrib., vol. 12, no. 4, pp. 842–849,
‘‘Distributed coordinated active and reactive power control of wind farms Feb. 2018, doi: 10.1049/iet-gtd.2016.2086.
based on model predictive control,’’ Int. J. Electr. Power Energy Syst., [52] D. Schulz, R. Hanitsch, T. Kompa, and A. Samour, ‘‘Comparative power
vol. 104, pp. 78–88, Jan. 2019, doi: 10.1016/j.ijepes.2018.06.043. quality investigations of variable speed wind energy converters with
[32] T. Ilakkia and G. Vijayagowri, ‘‘Hybrid PV/wind system for reduction of doubly-fed induction and synchronous generator,’’ in Proc. Power Quality
harmonics using artificial intelligence technique,’’ in Proc. IEEE-Int. Conf. Conf., D. Schulz, R. Hanitsch, T. Kompa, A. Samour, Eds., Aug. 2015,
Adv. Eng., Sci. Manage. (ICAESM), Mar. 2012, pp. 303–308. pp. 39–44.
[33] H. Ali, G. Magdy, B. Li, G. Shabib, A. A. Elbaset, D. Xu, and Y. Mitani, [53] I. A. Ahmed and A. F. Zobaa, ‘‘Comparative power quality study of
‘‘A new frequency control strategy in an islanded microgrid using virtual variable speed wind turbines,’’ Int. J. Energy Convers. (IRECON), vol. 4,
inertia control-based coefficient diagram method,’’ IEEE Access, vol. 7, no. 4, p. 97, Jul. 2016, doi: 10.15866/irecon.v4i4.10828.
pp. 16979–16990, 2019, doi: 10.1109/ACCESS.2019.2894840. [54] S. Agatonovic-Kustrin and R. Beresford, ‘‘Basic concepts of artificial
[34] H. Benbouhenni and N. Bizon, ‘‘Improved rotor flux and torque control neural network (ANN) modeling and its application in pharmaceutical
based on the third-order sliding mode scheme applied to the asynchronous research,’’ J. Pharmaceutical Biomed. Anal., vol. 22, no. 5, pp. 717–727,
generator for the single-rotor wind turbine,’’ Mathematics, vol. 9, no. 18, Jun. 2000, doi: 10.1016/s0731-7085(99)00272-1.
p. 2297, Sep. 2021, doi: 10.3390/math9182297. [55] M. Dawei, Z. Yu, Z. Meilan, and N. Risha, ‘‘Intelligent fuzzy energy
management research for a uniaxial parallel hybrid electric vehi-
[35] Y. Chongjarearn and P. Laohachai, ‘‘Stator-voltage vector control of a
cle,’’ Comput. Electr. Eng., vol. 58, pp. 447–464, Feb. 2017, doi:
doubly-fed induction generator,’’ in Proc. 34th Elect. Eng. Conf. (EECON-
10.1016/j.compeleceng.2016.03.014.
34), Chon Buri, Thailand, Nov. 2011, pp. 1–4.
[56] S. Soua, P. Van Lieshout, A. Perera, T.-H. Gan, and B. Bridge, ‘‘Deter-
[36] A. K. Agrawal, B. Munshi, and S. Kayal, ‘‘Study of wind turbine driven mination of the combined vibrational and acoustic emission signature of
dfig using Ac/Dc/Ac converter,’’ Tech. Rep., 2006. a wind turbine gearbox and generator shaft in service as a pre-requisite
[37] V. M. Gopala and Y. P. Obulesu, ‘‘A new hybrid artificial neu- for effective condition monitoring,’’ Renew. Energy, vol. 51, pp. 175–181,
ral network based control of doubly fed induction generator,’’ Int. J. Mar. 2013, doi: 10.1016/j.renene.2012.07.004.
Electr. Comput. Eng. (IJECE), vol. 5, no. 3, p. 379, Jun. 2015, doi:
10.11591/ijece.v5i3.pp379-390.
[38] G. Tapia, A. Tapia, and J. X. Ostolaza, ‘‘Proportional–integral regulator- RAMESH KUMAR BEHARA is currently pur-
based approach to wind farm reactive power management for sec- suing the Ph.D. degree in electrical power engi-
ondary voltage control,’’ IEEE Trans. Energy Convers., vol. 22, no. 2, neering with the School of Engineering, University
pp. 488–498, Jun. 2007, doi: 10.1109/TEC.2005.858058. of KwaZulu-Natal, Durban, South Africa. He is
[39] B. Desalegn, D. Gebeyehu, and B. Tamrat, ‘‘Evaluating the performances also an Electrical Engineer with Mondi Papers.
of PI controller (2DOF) under linear and nonlinear operations of DFIG- His research interests include advances in power
based WECS: A simulation study,’’ Heliyon, vol. 8, no. 12, Dec. 2022, systems and renewable energies control systems.
Art. no. e11912, doi: 10.1016/j.heliyon.2022.e11912. He is also a registered Professional Engineer-
[40] J. L. Rodriguez-Amenedo, S. Arnalte, and J. C. Burgos, ‘‘Automatic ing Technologist with the Engineering Council of
generation control of a wind farm with variable speed wind turbines,’’ South Africa and a member of SAIEE.
IEEE Trans. Energy Convers., vol. 17, no. 2, pp. 279–284, Jun. 2002, doi:
10.1109/TEC.2002.1009481.
[41] C. B. Moorthy and M. K. Deshmukh, ‘‘Wind resource assessment AKSHAY KUMAR SAHA (Senior Member, IEEE)
using computer simulation tool: A case study,’’ Energy Proc., vol. 100, is currently a Professor and an Academic Leader
pp. 141–148, Nov. 2016, doi: 10.1016/j.egypro.2016.10.156. Research and Higher Degrees with the School
[42] L. Tenghiri, Y. Khalil, F. Abdi, and A. Bentamy, ‘‘Optimum design of of Engineering, University of KwaZulu-Natal,
a small wind turbine blade for maximum power production,’’ IOP Conf. Durban, South Africa. His research interest
Ser. Earth Environ. Sci., vol. 161, no. 1, 2018, Art. no. 012008, doi: includes advances of power systems in various
10.1088/1755-1315/161/1/012008. areas, including engineering education. He has
[43] M. Ragheb and A. M. Ragheb, ‘‘Wind turbines theory—The Betz equation published more than 60 articles in top-tier inter-
and optimal rotor tip speed ratio,’’ Intechopen, vol. 1, no. 1, pp. 1–22, 2011. national journals and over 100 international con-
[44] P. Deka, ‘‘Influence of design parameters on the performance of Savonius ference papers in relevant areas. He is also
wind turbine,’’ Artic. Int. J. Innov. Sci. Technol., vol. 7, no. 4, pp. 121–130, a registered Professional Engineer with the Engineering Council of
2022. [Online]. Available: www.ijisrt.com South Africa and a fellow of the South African Institute of Electrical Engi-
[45] J. N. Libii, ‘‘Comparing the calculated coefficients of performance of a neers, South African Academy of Engineering. He is also a Senior Member
class of wind turbines that produce power between 330 kw and 7,500 kw,’’ of SAIMC, a member of Academy of Science of South Africa (ASSAf), and
World Trans. Eng. Technol. Educ., vol. 11, no. 1, pp. 36–40, 2013. an Individual Member of Cigre. He was awarded the Best Lecturer Electrical
[46] T. Ackermann, Wind Power in Power Systems. Hoboken, NJ, USA: Wiley, Engineering, in 2013 and 2014, (2016–2019) by the School of Engineering,
2005. Research Excellence Award, in 2015 and 2022, and Top-30 Publishing
[47] C. Hamon, K. Elkington, and M. Ghandhari, ‘‘Doubly-fed induction gen- Research Award for 2020 and 2021 by the University of KwaZulu-Natal.
erator modeling and control in DigSilent PowerFactory,’’ in Proc. Int. He is also acting as an editorial board member of a number of top-tier
Conf. Power Syst. Technol., Oct. 2010, pp. 1–7, doi: 10.1109/POWER- international journals. He is also a member of IEEE-HKN.
CON.2010.5666596.

38334 VOLUME 12, 2024