Wind Turbine Emulators—A Review

Abstract

Renewable energy sources have become a significant alternative energy source due to the continuing depletion of conventional energy sources and fluctuation in fuel costs. Currently, wind energy is the foremost among all other renewable energy sources. However, modeling and analyzing industrial wind turbines is complex as the wind turbine power ratio and size have steadily increased. Undoubtedly, industrial wind turbines are huge and challenging to keep in research labs; simultaneously, exploring the controller/power converter performance is practically impossible. Therefore, to overcome the above drawbacks, wind turbine emulators have been developed to achieve the static and dynamic characteristics of wind energy conversion systems. This paper aims to present a comprehensive review of the different wind turbine emulators available in the literature. In addition, the implementation of real-time emulators is classified according to the structure and approaches. Furthermore, an extensive analysis of the emulators was presented based on the significant parameters utilized for the real-time wind turbine emulators. Finally, this review analyzes the different emulator topologies according to cost, accuracy, complexity, and hardware implementation.

1. Introduction

The demand for electrical energy has increased worldwide and is associated with the environmental impacts caused by fossil fuels. Therefore, the generation of electricity based on renewable energy sources has become a more promising technology [1,2]. Among renewable energy sources, wind energy is the predominant renewable energy source due to its installed capacity worldwide. Furthermore, the technical merits of wind turbines (WTs) have led the increased generation capacities with a reduction in generation costs [3,4]. According to the report of the Global Wind Energy Council (GWEC), the total installed capacity of wind energy was approximately 837 GW in 2021 [5]. It shows that the significance of wind energy has continuously increased.

However, the real-time experiment on the physical wind turbine (PWT) system becomes complicated for the following reasons [6]: (1) large-area requirements; (2) expensive; and (3) discontinuous wind flow. In addition, further improvements in wind power systems demand different control schemes and power converter topologies for maximum power extraction and harmonic-free output. Therefore, testing the different control schemes and power converter topologies in the PWT is highly complicated; simultaneously, promoting research and education in the wind energy sector requires a wind turbine emulator (WTE). A WTE can mimic the characteristics of a large inertia machine using a small inertia machine without depending on the wind in a laboratory environment. The WTE is an alternative tool to test the performance of new control schemes and power converter topologies in the design and developmental stages. Furthermore, the hardware of the WTE replicates the dynamics of wind energy conversion systems (WECSs), and this emulator can analyze the static and dynamic behavior of the WT without wind flow and commercialized WT. In short, this emulator could reproduce the same torque and power that the PWT produces for a specific wind speed [7]. The physical implementation of the WTE consists of a motor and a generator, where the motor acts as the prime mover and generator for electrical subsystems. In general, the WTEs are designed and developed in the laboratory environment, which enables the flexible test bench to mimic the characteristics of physical WECSs. In addition, the emulators provide a platform for researchers, engineers, and designers to test and analyze the operation of WECS without depending on natural resources. Therefore, the essential requirement of WTEs is to develop the reference signal from the characteristics of the WT. There are two ways to generate the reference signal. Firstly, generate the reference speed from the mathematical model of the WT and utilize the control system to track the reference speed. Second, generate the reference torque by adopting the WT inertia and torque compensation schemes, and the motor should track the reference torque by adapting the different control schemes.

In the literature, various scholars have studied different methodologies for designing WTEs. In this paper, the authors have attempted to classify and summarize the scattered WTEs in the literature. In [8], the modeling of a wind turbine simulator (WTS) is classified into three categories according to its utility. The categories are as follows: (i) a simulator with the general structure; (ii) a simulator with dynamic behavior; and (iii) a simulator without dynamic behavior. In [9], the WTE has a separately excited DC motor that acts as a prime mover, and the microcomputer controls the DC motor. A dual DSP processor implements the PWT characteristics in the DC motor [7]. However, these methods did not include inertia or other compensation schemes. In [10,11], a reference current-based wind emulator was developed. However, these methods diminish the attributes of PWT. Therefore, a few modifications in the turbine model with the torque reference-based control of wind emulators are presented in [12,13,14,15,16,17,18]. The authors in [15,16,17] built the WTE with the effect of a tower shadow and inertia compensation at various wind speeds. In [19], a sensorless torque compensation was developed for a wind emulator, which reduces the utilization of the torque transducer. In [20], a power-based hardware setup of the wind emulator developed with the integration of large-scale wind energy is described. In Refs. [21,22], an open-loop WTE is designed without any feedback signals. These methods did not consider any feedback signal to emulate the characteristics of the WT. Polynomial expressions are used to approximate the power curve of the WT in [23,24,25]. Various control schemes have been proposed to implement static and dynamic characteristics in WTEs in [26,27,28,29,30,31,32,33,34].

A power reference-based WTE was developed in [35,36] to test and evaluate the PWT drive trains, which include generator and power electronic converters and controllers. However, this emulator does not have a power limitation delivered by the pitch control in variable-speed wind turbines (VSWTs). In [37], a Danfoss-VLT 5022 drive based on an open-loop torque control of the WTE was designed. Nonetheless, this system could not accurately emulate the impacts of tower shadow or wind shear. In [38], an emulator comprised an induction motor with a direct torque drive (ABB ACS 800). Furthermore, an observer was utilized to estimate the derivative of the angular velocity or speed with noise rejection. In [39], the authors considered the aerodynamic effect and torque compensation scheme to develop the WTE. Reference [40] presented the WTE with a low-cost and high-performance digital signal controller. The experimental results show that the simulator performs adequately under wind turbulence and the tower shadow effect. The transient stability and transient response of the emulator depend on the shaft of the WT. In [41], the WT dynamical model was derived from the three-mass model of the mechanical shaft, and a torque compensation scheme was used for emulating the WT characteristics. Most WTEs employ the torque compensation scheme to account for inertia compensation. However, this scheme suffers from noise and instability issues. Therefore, the one-degree-of-freedom (DOF) and two-DOF control structures proposed in [42] eliminate the above issues. The authors in [43,44,45,46,47,48,49,50,51,52,53,54,55] developed WTEs based on torque control schemes. The method described in [46] uses blade momentum theory to develop WTE. However, this study is more focused on blade design than on inertia or torque compensation. A platform for WTE was developed with a coupling of the two induction motors to emulate the characteristics of PWT [56]. In addition, the emulator is examined by different maximum power point tracking (MPPT) techniques. Furthermore, an improved compensation scheme proposed in [57] was used to improve the stability of WTE. However, this method could not guarantee the accurate emulation of the WT characteristics at different wind speeds. In [58], a laboratory setup was designed for a simple WTE, and it allows implementing the different control strategies with an analysis of major components. The reference torque to the induction motor drive is obtained through power and wind speed to emulate the WT characteristics [59].

In [60], fatigue, aerodynamics, structures, and turbulence (FAST) are utilized to model the aerodynamic and mechanical aspects of the wind turbine. Additionally, this emulator can investigate the yaw effect, which is more severe than the tower shadow. Similarly, in [61], the vertical and horizontal effects of wind shear are modeled in the proposed emulator. In [62], the authors addressed the significance of the $C_p$–$\lambda$ curve for micro wind turbine operation and designed the emulator for a small wind turbine with the development of different control schemes. Field-oriented control (FOC) approaches such as indirect FOC [63] and fuzzy-based indirect FOC [64] were proposed to enhance the dynamics of the WTE. A low-cost intelligent relay-based WTE was developed, and the control algorithms were implemented in the intelligent relay using the function keys [65]. However, this method does not consider the effects of aero turbine and mechanical loads on the WT. In [66,67], commercial induction motor drives such as the Danfoss drive and the Parker AC drive emulate the dynamics of the wind turbine. In the literature, few WTEs use a doubly fed induction motor [68], servo motor [69,70], and permanent magnet synchronous motor [71,72,73] as a prime mover for wind emulators. Furthermore, the authors in [74,75,76,77] presented a detailed study on computational WTEs. According to the literature, most WTEs are under the category of a combination of physical and computational emulators [7,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,41,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,78] and a few emulators are in the category of pure computational emulators [74,75,76,77].

In PWTs, exploring the control schemes and power converter topologies is impossible. Therefore, WTEs are mainly used for the above purposes. In addition, the WTEs are used in a microgrid to test the performance of hybrid power systems. For instance, the authors in [79] proposed an induction motor-based WTE capable of correlating the static and dynamic characteristics of PWT and WTE. In addition, a microgrid was formed by WTE integrated with a photovoltaic (PV) and battery storage system. Furthermore, the performance of the microgrid was investigated with three different scenarios. In [80], a hybrid wind power system was developed by designing the PV generator and WT simulator with batteries. This test bench allows the realistic emulation of multiple renewable energy sources with appropriate control strategies. A microgrid laboratory was developed in [81] by integrating the PV emulator, WTE, and batteries. It also investigates the difficulties and constraints associated with interconnecting renewable sources.

From the literature, it is clear that extensive research has been conducted on WTEs. However, a comprehensive review has not yet been presented in this research area. In this context, research on WTEs significantly requires a review article that describes the various design aspects and their applications. It will be more beneficial for academics and researchers to choose an appropriate design for the WTE. This review work has the following contributions.

• This work presents an overview of WTEs and summarizes the different WTEs available in the literature. To the best of our knowledge, no previous research has been investigated on reviewing WTEs in detail.
• Detailed classification and investigation were conducted for various WTEs. Furthermore, this review presents a comparative analysis of different methodologies adapted for WTEs based on cost, accuracy, and complexity.
• Furthermore, an extensive analysis of the different WTEs, such as a combination of physical and computation emulators and pure computation emulators, are investigated and illustrated based on parameters such as drive train dynamics, tower shadow, wind shear, $C_p-\lambda$ curve, torque and inertia compensation, hardware, and converter topologies.
• Finally, a comparative study was performed based on the following parameters, such as: (1) realization cost; (2) accuracy of the emulator; (3) level of complexity; and (4) hardware implementation.

The organization of this paper follows. Section 2 describes the classification of wind turbine emulators. Section 3 and Section 4 present a discussion of the combination of physical and computational WTE and pure computational emulators, respectively. Section 5 examines the detailed comparative study of different WTEs. Finally, Section 6 illustrates the conclusion and future recommendations.

2. Classification of Wind Turbine Emulator

Real-time (RT) emulators are mandatory for mimicking the behavior of any physical system where the physical system demands a higher cost and space management. For instance, RT emulators play a significant role in WECS to emulate the dynamic characteristics and controller evaluation. Figure 1a shows the basic schematic representation of WECS. The essential components of the WECSs are the wind turbine, generator, power converters and rotor and grid-side controllers. Nowadays, variable speed WECS are widely used in the wind industry with direct drive or reduced gearbox. Figure 1b shows the representation of the WTE (combination of the physical and computational emulator). An emulator is an essential hardware to analyze the dynamics of the WECS without wind flow and actual wind turbines. According to the literature, WTE is classified mainly into two categories [77]: (1) combination of a physical and computational emulator; and (2) computation emulator.

2.1. Combination of a Physical and Computational Emulator

This approach is most frequently utilized in the development of WTEs. The emulator is divided into two major parts: WT electrical and WT mechanical subsystems. Electrical generators have developed the emulation of electrical subsystems with power circuitry, whereas AC or DC motors are employed to emulate the WT mechanical subsystems. In addition, the mechanical torque is calculated via WT mathematical model, and that can be implemented on the AC or DC motor using controller boards.

2.2. Pure Computation Emulator

The real-time WTE was developed as a software on a computational device or with specialized simulators such as RT-LAB, LabVIEW, and Simulink [74,75,76,77]. The significant merits of this approach are its cost-effectiveness and high flexibility.

3. Classification of Combination of Physical and Computational

The combination of physical and computation emulators has been further classified according to the utilization of motors. Generally, three types of motors are used for WTEs: direct current (DC) motor; induction motor (IM); and synchronous motor. The subsequent sections will review the existing WTEs reported in the literature based on the types of electric motors. Figure 2 presents the detailed classification of a combination of the physical and computational emulators.

3.1. DC Motor Based WTEs

This section presents a detailed analysis of the various WTEs based on DC motors.

3.1.1. Reference Current

A microcomputer-based control of WTE is proposed in [9]. The dynamic model of the WT system consists of resistance torque, inertia, and viscous friction. Thus, the net torque ($T_r$) can be expressed in (1).

$$T_r=T_t-J_r\frac{d\omega }{dt}-B_r\omega -K_d-T_g$$

(1)

where $T_g$ is the generator torque, $T_t$ is the turbine torque, $T_r$ is the net torque, $J_r$ is the rotor inertia, and $\omega$ is the rotor speed.

Equation (2) expresses the generator torque ($T_g$) derived from the voltage across the load (V) and the current through the load ($I_L$).

$$T_g=\frac{V·I_L}{\omega ·{\eta }_g}$$

(2)

$$\frac{d\omega }{dt}=\frac{T_r}{J_r}$$

(3)

In addition, the reference speed is derived using the relationship given in Equation (3). Furthermore, the reference current is generated by the difference between the reference and the measured speed. Finally, a thyristor bridge was employed to control the DC motor according to the variation in wind speed.

In [7], a separately excited DC motor acts as a WTE, and the control of the emulators is achieved by dual digital signal processing (DSP) units. The DSP UNIT I is dedicated to sampling the shaft speed. In DSP UNIT II, the reference current is generated, taking into account the following parameters: motor speed, wind speed sequence, and torque speed characteristics ($\frac{T}{\omega }$). Finally, a current controller is established to drive the DC motor according to the error between the reference current and the measured current.

In [10], WECS was implemented through DFIG coupled to a DC motor. A DC motor emulates the characteristics of the WT and is controlled by the reference armature current. This current is a function of the motor torque and is equal to the turbine torque generated by the wind speed in a steady-state model. The reference armature current is calculated using Equation (4) and compared with the measured armature current. A PI controller generates the PWM pulse to the chopper to generate the turbine characteristics in a DC motor.

$$I_{DCM}^{*}=\frac{\rho A}{2\ast G\ast K_{DCM}\ast I_{ex}}.\frac{C_p{\upsilon }^3}{{\mathrm{\Omega }}_m}$$

(4)

where $I_{ex}$ is the excitation current, $C_p$ is power coefficient, $\rho$ is the air density, A is the area of the rotor, $\upsilon$ is the wind speed, ${\mathrm{\Omega }}_m$ is the rotor speed, and $K_{DCM}$ is the electromotive force coefficient.

Most wind emulators have utilized speed or torque sensors to emulate the characteristics of the WT. However, to overcome the above drawback, a sliding second order based on a super-twisting observer was proposed in [11] to estimate the reference current for WTE. In addition, a fuzzy logic controller is utilized to ensure the closed-loop operation of the emulator. Finally, the proposed methodology was experimentally verified through dSPACE.

3.1.2. Reference Torque and Torque Compensation Based

A 4.4 kW DC motor emulates the characteristics of the WT along the commercial drive with torque control [12]. In addition, a reference torque is calculated using Equation (5) by a control program on the dSPACE control board, which includes the tower shadow ($T_{ts}$) and the wind shear effect ($T_{ws}$). The control program reads the wind velocity from the input file and the rotor speed from the encoder. Finally, the reference torque is transferred as a voltage signal to the DC drive to achieve the steady-state characteristics of the WT.

$$\begin{array}{c}\hfill T_{ts}=-t_{ts}·cos\left(\Psi \right)\\ \hfill \frac{V_1}{V_2}={\left(\frac{h_1}{h_2}\right)}^a\\ \hfill T_{ws}=-t_{ws}·cos\left(\Psi \right)\end{array}$$

(5)

where $\Psi$ is the angle blades to the tower, $t_{ts}$ and $t_{ws}$ are the empirical coefficient, and a is the coefficient for terrain roughness.

In [13], a laboratory-scale WTE is designed with a 7.5 kW DC motor coupled with an induction/synchronous permanent magnet generator. The thyristor converter controls the DC motor and acts as a prime mover. The aerodynamic characteristics of a rotor blade are a function of wind speed and rotational speed. In addition, the tower effect is considered in terms of the periodic pulsation torque. An additional inertia compensation is prepared by a low-pass filter and a derivative of the rotation speed. This compensation torque represents the inertia difference between the motor and the blade. The proposed algorithm calculates the reference torque on the basics of inertia of the rotor blade. However, the low control bandwidth limits the performance of the WTE and introduces high torque ripples. The dynamics of the drive train and the inertia of the wind turbine are considered for the development of WTE [14,15,16]. The harmonic torque due to the gradient and tower shadow effect is also considered and is given by (6)

$$T_{WT}=T_{AV}\left[1+Asin\left({\omega }_{WT}t\right)+Bsin\left(3{\omega }_{WT}t\right)\right]$$

(6)

where $T_{WT}$ and $T_{AV}$ are the aerodynamic and average torque, respectively, and ${\omega }_{WT}t$ is rotational shaft speed. The WTS develops the reference torque, consisting of aerodynamic torque, fluctuation torque due to wind shear and tower shadow, and compensation torque due to large inertia. The PI controller regulates the single-phase half-controlled converter based on the difference between the reference and actual torque and is implemented in the dSPACE control board.

A physical simulator for WECS was presented in [17]. It has a simulator, dedicated control algorithms, and the dynamics of a real WT model. The hardware in the loop (HIL) simulation is adapted for developing WECS, and it has two parts, namely a real-time physical simulator (RTPS) and investigated physical systems (IPSs). The RTPS represents the aerodynamic and mechanical transmission of wind turbine systems, whereas the IPS represents electrical systems that exist in real time. The interaction between RTPS and IPS replicates the dynamic behavior of WECS. First, the turbine torque is derived using the relationship in (7). Then, the dynamics of the rotor speed in the HIL simulator are represented in (8). The reference torque is derived from (9), and is converted to the reference current. Finally, the reference current is compared to the actual current, and the error is passed through a PID controller to generate the PWM signal to the chopper.

$${\mathrm{\Gamma }}_{WT}\left({\mathrm{\Omega }}_L,\upsilon \right)=0.5\rho \pi {\upsilon }^2{R}^3{C}_{\mathrm{\Gamma }}\left(\lambda \right)+{\mathrm{\Gamma }}_{ts}-{\mathrm{\Gamma }}_{fv}-{\mathrm{\Gamma }}_s$$

(7)

$$J_E\dot{{\mathrm{\Omega }}_R}={\mathrm{\Gamma }}_E-{\mathrm{\Gamma }}_G\left({\mathrm{\Omega }}_R\right)$$

(8)

$${\mathrm{\Gamma }}_E^{*}=\frac{\eta {\mathrm{\Gamma }}_{WT}}{i}-\left(\frac{J_{wt}\eta }{i^2}-J_{DCM}\right)\dot{{\mathrm{\Omega }}_R}$$

(9)

where ${\mathrm{\Gamma }}_{ts}$, ${\mathrm{\Gamma }}_{fv}$, and ${\mathrm{\Gamma }}_s$ are the torque generated by the tower shadow effect, viscous friction torque, and the static friction torque, respectively. ${\mathrm{\Gamma }}_{WT}$ is the wind torque, ${\mathrm{\Gamma }}_E^{*}$ is the reference torque, and ${\mathrm{\Gamma }}_E$ electromagnetic torque.

The dynamics of the 5 MW direct drive wind turbines are emulated by the 300 W separately excited DC motor and controlled by the current-controlled DC drive [18]. The current of the DC motor is controlled by a three-phase semi-controlled rectifier bridge. To emulate the 5 MW turbine, the emulator is divided into two parts, that is, the steady-state and transient characteristics. The steady-state characteristics are obtained by implementing the real wind speed profile on the DSP board similarly to the wind profile generated by Turbsim. However, the transient characteristics of the 5 MW turbine are not identified as a 300 W DC motor. Furthermore, the inertia of a 5 MW turbine is added to the system to overcome this issue. Finally, the wind emulator is experimentally validated by different stochastic wind profiles.

3.1.3. Observer Based

The authors in [19] have proposed a torque sensorless inertia compensation algorithm for a real-time WTS with larger or smaller inertia. An incremental encoder measures the relative position angle of the motor. In addition, an observer is employed to estimate the speed and acceleration, as depicted in Figure 3. The generator torque is estimated from the torque estimator using the derivative of the rotor speed. Equation (10) describes the reference torque of the DC motor and the current reference obtained from this torque. This proposed algorithm is verified with the hardware set-up for 2 kW WTS with a digital controller.

$$T_M=\frac{J_M+J_G}{J_T+J_G}{T}_W+\frac{J_T-J_M}{J_T+J_G}{T}_G+T_F$$

(10)

where $T_M$ is DC motor output torque, $T_F$ is the equivalent torque of total mechanical torque, and $T_G$ and $T_W$ are the electromagnetic and turbine output torque, respectively. $J_M$, $J_G$, and $J_T$ are the inertia of the DC motor, generator, and turbine rotor, respectively.

3.1.4. Power Based

The design and development of a wind turbine test platform consisting of a DC motor with a 1 kW grid-coupled synchronous generator was presented in [20] and illustrated in Figure 4. The power is calculated from the generator voltage and current and then compared with the reference power. The power error is generated and sent to a PI regulator to generate the reference speed. Finally, the reference speed is compared with the measured speed, and then the error signal sends to the PI controller to generate the reference current.

3.1.5. Open Loop

The authors in [21] presented a new WTE that can simulate turbine power curves without a closed-loop control system. This emulator comprises the following components: a DC motor with independent excitation, a DC voltage source, and power resistors. The aim was to develop a DC motor power curve which is equivalent to the WT power curve at a given wind speed. The mechanical power of the DC motor is the product of the electromotive force and motor current. Equation (11) presents the mechanical power and the maximum power is obtained by differentiating this equation with respect to the motor speed at ${\omega }_{max}$.

$$P_{\mathrm{m}}=\frac{k{I}_{\mathrm{f}}}{R+R_{\mathrm{m}}}\left(V_{\mathrm{DC}}\omega -k{I}_{\mathrm{f}}{\omega }^2\right)$$

(11)

where $P_{\mathrm{m}}$ is the mechanical power of the DC motor, $\omega$ is the rotor speed, $V_{DC}$ is the DC voltage source, k is the constant, $I_f$ is the field current, R is the series resistance, and $R_m$ is the stator resistance. Additionally, the variation in wind speed is achieved by varying the DC voltage. The main advantage of this method is the absence of feedback signals and online calculations. The tuning of this emulator could be achieved by varying the parameters such as the DC voltage, series resistor, and field current. However, the series resistor is fixed because it cannot be changed online. Finally, the wind variations were achieved by adjusting the DC voltage and field current. Therefore, the power curves developed by this emulator are not identical to actual power curves. In addition, an independent adjustment of the tuning parameters is essential for every wind variation.

In [22], a 300 W open-loop WTE was developed using MATLAB/XSG. Initially, the wind velocity was converted into the rotation speed using the mathematical model of the wind turbine and the $C_p\left(\lambda \right)$ curve. Furthermore, a duty cycle calculator was employed to obtain the duty ratio, which is sent to the buck converter. However, this conversion may not be helpful for the realization of real turbine characteristics.

3.1.6. Polynomial Approximation

A separately excited DC motor is employed to develop a small isolated WTE and the performance of the emulator is investigated under different control strategies [23,24]. The approximate $C_p\left(\lambda \right)$ model is presented in Equation (12), and the $R^2$ for this model is approximately 99.8%. Equation (13) gives the relationship between the wind speed and furling angle ($\Theta$) and the $R^2$ for this model is approximately 98.59%. In the case of small WTs, the effective wind is modeled as $Vcos\Theta$. Equation (14) represents the reference rotor speed, and the torque information mentioned in this equation is obtained from the motor current. However, these kinds of approximations on the WT power curve require a dataset, which is only feasible for small WTs.

$$C_p\left(\lambda \right)=0.00044{\lambda }^4-0.012{\lambda }^3+0.097{\lambda }^2-0.2\lambda +0.11$$

(12)

$$\Theta =-0.0001732V^5+0.0085008V^4-0.1203V^3+0.4501V^2+1.0592V+0.38972$$

(13)

$${\omega }_m=\frac{0.5\rho A{C}_p\left(\lambda \right){\left(Vcos\Theta \right)}^3}{T_w}$$

(14)

Different design requirements and different control schemes are discussed in [25]. The static and dynamic characteristics of WTE are obtained from a look-up table or a polynomial expression. The $C_p\left(\lambda \right)$ curve is represented by a 6th order polynomial using Equation (15) and the reference torque (T) is derived from it. The reference torque is converted into the reference speed using (16). The current reference is obtained from the speed control loop, and then the DC motor emulates the real characteristics of the WT.

$$C_p\left(\lambda \right)=a_0+\sum _{i=1}^{i=6}{a}_i{\lambda }^i$$

(15)

$$dn=\frac{Tdt}{2\pi J}$$

(16)

3.1.7. Conventional PI Controller

This section illustrates the emulation of an actual wind turbine by a DC motor and PI controller. Generally, the reference to the PI controller is classified into three categories: torque reference, current reference, and torque and current reference. In the current reference mode [26,27], the reference current is derived from a mathematical model of the wind turbine. The difference between the reference current and the armature current is passed through the PI controller to generate PWM pulses. In speed and current reference mode [28,29], the reference speed is compared with the actual speed, and then the error is sent to the PI controller for generating the reference current to the inner loop PI controller. Then, the inner loop PI controller generates the pulses for the DC motor drive. In torque reference mode [30], the reference torque is obtained from the generator power and compared to the actual torque of the motor. Then, the PI controller generates the pulses based on the difference between the two torques. A DC motor-based WTE is developed, and the black widow optimization algorithm (BWOA) tunes the control parameters such as the proportional and integral gains [31,32]. A reference WT model acquires the wind speed, pitch angle, and motor speed and produces the reference current. Finally, the reference current is compared with the actual current and passes through the BWOA-tuned PI controller. In [33], a real-time WTE with a robust sliding mode controller was proposed. A reference current is generated by the static and dynamic characteristics of the WT model, and the sliding mode theory is utilized to develop a control law to track the reference current.

3.1.8. Commercially Available Wind Turbine Emulators

An ECOSENSE WTE mimics the actual wind turbine in hardware-level simulations [34]. It has a DC motor coupled with an induction generator, and the speed of the motor is controlled by computing the reference speed, which is derived from the mathematical model of the wind turbine. This emulator has several possibilities to implement the newly designed or modified wind turbine mathematical model to emulate the power/speed characteristics in the hardware environment for different wind speeds and pitch angles.

Table 1. Comparative analysis of different WTEs based on DC motor.

S. No	References	Variable Required to Build an Emulator (Motor)	Dynamics Consideration for Emulator					Converter and Control Technique for Emulator	Hardware Implementation	Emulator Accuracy
			${\mathit{C}}_{\mathit{p}}-\mathit{\lambda }$ Curve	Tower Shadow Effect	Wind Shear Effect	Torque Compensation	Inertia Compensation
1	[9]	Speed	✓					Armature voltage control	Microcomputer	*
2	[7]	Speed	✓					Current control	DSP	*
3	[10]	Speed	✓					DC–DC converter	dSPACE DS1104	*
4	[11]	Speed	✓					DC–DC converter	dSPACE DS1104	*
5	[12]	Speed	✓	✓	✓			DC drive with torque control	dSPACE	**
6	[13]	Speed		✓			✓	PWM IGBT vector control	DSP	***
7	[14]	Speed		✓			✓	DC–DC converter	DSP	***
8	[15,16]	Motor current		✓	✓		✓	Single phase half-controlled DC drive with torque control	MATLAB/Simulink real-time control software interfaced with I/O board	****
9	[17] ${}^a$	Speed	✓	✓				Control rectifier and current control	DSP TMS320F28335	***
10	[18]	Speed	✓			✓	✓	3-phase semi-controlled rectifier	DSP TMS320F28335	**
11	[19]	Position	✓				✓	DC–DC converter with current controller	DSP and FPGA board	**
12	[20]	Speed	✓					Thyristor controlled armature current	DSP	*
13	[21]	Speed	✓					Thyristor rectifier	Variable speed drive from Schneider Electric (ATV312HU15M3)	**
14	[22]	Speed	✓					DC–DC converter	FPGA board	*
15	[23,24]	Speed	✓ (Furling angle)					DC–DC converter	Lab Master I/O board	*
16	[25]	Speed	✓					Thyristor controlled current controller regulator	DSP TMS 320LF2407	*
17	[26]	Speed	✓					Armature voltage regulator	LabVIEW	*
18	[27]	Speed	✓					H bridge DC–DC converter	TMS 320LF28335	*
19	[28]	Speed and Current	✓					DC–DC converter	dSPACE DS1104	*
20	[30]	Speed	✓					DC drive	Arduino uno	*
21	[31]	Speed	✓					DC–DC converter	Arduino mega	*
22	[32]	Speed	✓					DC–DC converter	dSPACE DS1104	*
23	[33]	Speed	✓					DC–DC Converter	MATLAB/Simulink	*

^a Static and viscous friction included, * Less, ** Moderate, *** High, and **** Very high.

shows a detailed comparative analysis of various WTEs based on DC motors.

3.2. Induction Motor-Based WTEs

This section illustrates an exhaustive analysis of different WTEs based on induction motors.

3.2.1. Power Reference Based

The WTS was designed for WECS to simulate the characteristics of an actual wind turbine. It consists of an induction motor with a power rating of 10 hp, and drives the low-speed generator (10 kW) using a variable-speed drive [35,36]. Based on the model of the induction motor, the stator current and frequency demand were calculated by the control program for a specific power and speed. The torque/speed transducers measure the torque and speed as a feedback signal. The torque error is generated from the measured and emulated torque (obtained through power and speed) and is passed through a PI regulator to generate the reference current. An Intel 80C196KD microcontroller drives the IGBT inverter by comparing the output current from the PI regulator to the actual current and generating the triggering pulse by the current hysteresis control strategy. However, this scheme is expensive due to the utilization of the current sensor and various hardware and software protection. Figure 5 shows the power reference-based WTE.

3.2.2. Open-Loop Torque Control

A 15 kW induction motor acts as a mechanical shaft of the emulator, and the commercial frequency inverter controls the motor in an open loop [37]. The aerodynamic model of the WT is developed based on the $C_p(\lambda ,\beta )$ curve, and a reference torque is then calculated using (17). The frequency converter (Danfoss-VLT 5022 scalar control) receives input as reference torque ($T^{*}$) to control the induction motor in open-loop torque control mode. However, the accuracy of the emulator is highly dependent on the approximation of the $C_p(\lambda ,\beta )$ curve.

$$T^{*}=\frac{1}{2}\rho \pi R^3{\upsilon }^2{C}_Q$$

(17)

where $\rho$ is the air density, R radius of the rotor, $\upsilon$ wind speed, and $C_Q$ is the torque coefficient.

3.2.3. Observer-Based Torque Control

Direct torque control (DTC) is implemented on the dSPACE platform to develop an induction motor coupled with PMSG-based wind emulator in [38]. First, the WT torque is calculated from the WT model, which considers inertia compensation. The input to the WT models is the wind speed and the speed of the motor. Then, the closed-loop torque observer estimates the speed of the motor and torque. This estimated speed is compared with the actual speed and then the speed error is passed to the PI regulator to compute the torque reference given in Equation (18). Finally, the reference torque is obtained and sent to the DTC drive for controlling the induction motor. The implementation of the experimental design contains dSPACE and DTC inverter.

$$T_{IMref}=\frac{T_{wt}+T_{GB}}{\left(J_c+J_{IM}\right)\frac{d{\omega }_r}{dt}}$$

(18)

where the $T_{GB}$ gearbox friction torque, $T_{wt}$ wind turbine torque, $T_{IMref}$ reference torque, ${\omega }_r$ rotor speed, $J_c$ compensation inertia of the emulator, and $J_{IM}$ inertia of the induction motor.

3.2.4. Torque Compensation-Based Torque Control

The dynamic WTE in [39] was designed by including factors such as the wind shear, tower shadow, torque compensation scheme and presented in Figure 6. This scheme compensates for the inertia difference between the actual and laboratory machine. The dynamic torque is viewed as a reference torque for the control of an induction motor, and it is calculated using DSP. A vector control method controls an induction motor based on the reference torque. Equation (19) describes the compensation torque ($T_{\mathrm{comp}}$).

$$T_{\mathrm{comp}}=T_{\mathrm{blade}}-T_M=\left(J_B-J_M\right)\frac{d{\omega }_g}{dt}$$

(19)

An induction motor driven by a torque control inverter develops the WTS. It consists of wind speed simulation, the mathematical model of a WT, blade characteristics & inertia, and the tower shadow effect. The algorithm is implemented in a digital signal controller (DSC) [40,41,78]. First, it receives the rotation speed from the encoder and the wind speed from the source, and then the tip speed ratio, generator speed, and angular acceleration are calculated. Then, the lookup table computes the aerodynamic torque and the DSC calculates the torque compensation and torque ripple due to the tower effect. Finally, a compensated torque is developed and sent to a torque-controlled inverter to control an induction motor. A proposed wind simulator was experimentally verified by a 4 kW squirrel cage IM coupled with a 1 kW DC generator, and performed satisfactorily during turbulence and tower effect.

The authors in [42] addressed the torque compensation scheme for the WTE. Generally, a torque compensation scheme is utilized to compensate for the difference between the inertia of the WT and the motor. However, this compensation loop suffers from instability and noise. Therefore, to overcome the above limitations, one DOF (1-DOF) and two DOF (2-DOF) control structures for wind emulators are utilized. In 1-DOF, the main objective of the control is to track the actual speed, and the reference speed is obtained by taking the difference between the generator torque ($T_G$) and the turbine torque ($T_T$). Since real $T_T$ is not measured directly, it is estimated by stator and rotor currents. The disturbance rejection capability is poor in 1-DOF and can be solved by feed-forward compensation. The reference speed is obtained in 2-DOF by changing the $T_T$ and prefilter. The torque reference is the output of 2-DOF, which is translated into the reference current, and these references are fed to the controller.

3.2.5. Torque Control

The wind simulator is constructed by a 5.5 kW induction motor with a variable speed inverter, and the power speed characteristics of the WT were emulated by the Programmable Logic Controller (PLC) [43]. For different wind speeds, the torque–speed characteristics are embodied in the wind simulator with aerodynamic effects such as the wind turbulence, yaw error, and passive stall. The PLC generates the reference torque from the wind and rotational speeds and sends the torque to the ABB motor drive.

A laboratory setup WTE has been developed based on a frequency inverter with a direct torque control scheme [44]. The torque transducer measures speed and torque, and then the PIC32MX360F512L microcontroller calculates a reference torque using the wind speed and the rotor speed [44]. Finally, the reference torque is compared with the measured torque to emulate the IM as the emulator. However, the authors did not consider the inertia of the actual wind turbine during its implementation.

The wind turbine model was developed based on blade element moment theory [45,46,47,48]. It consists of three-dimensional wind turbine characteristics with a mechanical model. Wind velocity is an internal condition of the emulator and the rotor speed is measured by a rotary encoder (RE). The three-dimensional (3D) table computes the torque based on the wind speed and rotor speed. The transient and steady-state characteristics of the WTE strongly depend on the viscous coefficient and moment of inertia of the actual WT. Therefore, it is essential to compensate for these parameters because the wind speed changes continuously. The DSP/dSPACE calculates the reference torque ($T_{im}^{ref}$) from Equation (20), and this torque is sent to the IM via an off-the-shelf inverter.

$$T_{im}^{ref}=\frac{J_{im}}{J_t}\left(T_t-B_t{\omega }_t\right)+B_{im}{\omega }_t$$

(20)

where $J_{im}$ is the moment of inertia of the machine, $B_{im}$ is the coefficient of viscosity of the induction motor, ${\omega }_t$ is the rotor speed, $J_t$ is the moment of inertia of the WT, and $T_t$ is the torque of the WT. An induction motor with an IGBT inverter-fed squirrel cage acts as a WTS in [49]. In this method, a user can define the wind speed profile and the speed transducer measures the rotor speed. For instance, if the power and tip speed ratio ($\lambda$) are known for a constant wind speed, then the output power is only related to the rotational speed. The characteristics of power vs. speed are similar to the $C_p-\lambda$ curve, and the mechanical torque can be calculated through the $C_p-\lambda$ curve and stator current of IM. In addition, the relationship between $i_{qs}$ and the rotor speed is derived from the mechanical torque. Finally, the current control is implemented on the DSP controller based on the difference between the reference and measured q axis current. The controller sends the triggering pulse to the driver circuit to obtain the power vs. speed curve for the emulator.

A squirrel cage induction motor coupled with a DFIG-based wind emulator was discussed in [50]. The modeling WT characteristics depend on the power vs. wind speed and blade pitch angle. Additionally, this model includes the tower shadow effect and wind shear. The modeling WT characteristics depend on the power vs. wind speed and blade pitch angle. The reference torque is calculated from the turbine power and speed. In addition, the reference current is obtained by dividing the turbine torque by its torque constant. The proposed methodology was validated with the laboratory setup of a 2.3 kW IM-based WTE, and the torque controller drives the IGBT power converters to emulate the WT characteristics.

In [51], the static characteristics of a WT were emulated by considering the MATLAB dynamic model, SCADA, frequency converter, induction motor, and signal conditioning. The different spectrums of wind speed were considered to generate the wind speed, and the turbine torque was derived from the power of the turbine and the speed of the rotor. The difference between the turbine and generator torque gives the rotor speed. This speed was sent to the frequency converter via SCADA to emulate the WT characteristics in the IM. In this emulator, the PLC controls the yaw and pitch actuator by receiving the wind speed and direction as input to emulate the turbine in a different mode of operation. However, this study ignored the dynamic aspect of the WTE. The authors in [52] introduced the torque transducer and rotary encoder in a WTE to perform a closed-loop operation. These sensors measured signals, such as torque and speed signals, and sent them to the computational model. This model plots the power curve to obtain the set-point voltage for the VFD to control the IM. In addition, a LabVIEW was utilized to implement the control algorithm with NI 622. Finally, the emulator replicated the WT behavior by considering the fluctuations in wind speed, pitch angle, and load effects. In [53,54], the authors discussed the design of a three-phase IM-based WTE with the utilization of a frequency inverter. The closed-loop control utilized the torque sensor and an encoder to emulate the actual characteristics of the wind turbine. The power curve of the WTE was developed using these signals in the computational model, and the voltage level applied to the frequency inverter was also obtained. Finally, the control algorithm for the emulator was developed in a MATLAB environment and communicated to the test bench through the DSP TMS320F28335. In [55], a dead-beat current controller was employed in WTE. The reference torque for an IM was derived from wind power and kinetic energy. A state-space model of the current controller was derived and a deadbeat algorithm is used to find the L matrix. The transfer matrix decoupled the static and dynamic characteristics of the current components.

3.2.6. Improved Inertia Torque Calculation

In [56], two identical 1.5 kW induction machines were coupled together to emulate the static and dynamic characteristics of WECS. The calculations related to the emulator were performed on a PC and DSP, and both were communicated via USB. The final torque applied to the WT was obtained by the difference between the calculated torque on PC and the inertia torque calculated on DSP. The inertia torque calculation was improved by considering the shaft speed derivative and implemented in DSP with a 20 kHz sampling frequency. This implementation is relatively simple, but introduces high frequencies in generator speed acquisitions. At the same time, this method is not suitable for large turbine inertia. A low-pass filter eliminates the noise, but the noise is amplified due to a large gain. The discrete derivative with a first-order high-pass filter with Tustin transformation solves the above issue. With this structure, the shaft speed is noise-free. Figure 7 shows the block diagram of the measurement of inertia torque calculation using a high-pass filter [56].

In [57], a discrete linear model of WTS was implemented considering the time delay of the acceleration observer. An improved compensation scheme with a first-order digital filter was applied for large turbine inertia to moderate the deviation response. The implementation of the enhanced compensation scheme was achieved by adding the first-order digital filter to the torque compensation loop. As a result, the open-loop transfer function of WTS was modified. It has two components: the first was similar to the expected response, and the second was the new exponential deviation. Adjusting the filter coefficient minimizes the exponential deviation, and the WTS response is similar to that of a real wind turbine. Figure 8 shows a WTS with a first-order filter.

3.2.7. Lookup Table

A three-phase induction motor implements the power–speed characteristics of the actual wind turbine with the closed-loop control of the $\frac{\mathrm{v}}{f}$ method [58]. It has an outer speed control loop and a PI regulator, generating slip added to the shaft speed for obtaining the stator frequency. By using $\frac{\mathrm{v}}{f}$ control, the voltage command was generated at the same time this $\frac{\mathrm{v}}{f}$ ratio was maintained. Therefore, the torque of the motor only depends on the slip. The look-up is utilized for programming the machine terminal voltage. The PWM pulses to the IM drive are generated based on wind and rotor speeds. In addition, the speed signal was estimated from the terminal voltage and current using DSP/dSPACE. Figure 9 shows the torque estimation based on the lookup table.

3.2.8. Power and Wind Speed Relationship

In [59], an induction motor-based WTE was designed via ordinary commercial drive. The WT block considers the input as wind speed and rotor speed and generates the shaft torque as an output. In addition, the wind speed ranges were considered from 4 to 8 m/s with a fixed pitch angle. The linear relationship between power ($P_{ele}$) and wind speed ($\upsilon$) is expressed in Equation (21), and the reference torque ($T_{ref}$) is derived using Equation (22). A PIC16F876A microcontroller was utilized to implement this proposed methodology in real time. However, this proposed methodology significantly relies on the approximation of the power and wind speed and does not consider the essential dynamics of the WT. This method linearly relates the power and wind speed, despite these parameters being nonlinear.

$$P_{ele}=\left(0.3858\upsilon -1.3114\right)\ast 1000$$

(21)

$$T_{ref}=\frac{\left(0.3858\upsilon -1.3114\right)\ast 1000}{{\omega }_r}$$

(22)

3.2.9. FAST Simulator-Based Emulator

The WTE was developed in [60] based on the dynamics of different components of the WT, such as aerodynamic, mechanical, and electrical. This methodology was implemented by an induction motor and a generator set with a torque transducer and variable speed drive (VSD). The tower shadow and the yaw effect are considered in this proposed emulator for analyzing the power fluctuations and fatigue loads. The mechanical and aerodynamics of a two-bladed downwind WT were modeled in a LABVIEW environment using Aerodyn and FAST simulation tools. In addition, the electrical subsystems were implemented in real-time. The FAST code calculated the high-speed shaft (HSS) and sent it to the VSD to control the rotational speed of the IM-IG set. The torque transducer measures the torque and sends it to the FAST code through the DAQ card. The 275 kW wind turbine is considered to study the effect of tower shadow and yaw error. Thus, the measured torque is scaled to a 275 kW wind turbine. This model helps emulate the realistic behavior of WECS.

In [61], the modeling and analysis of horizontal and vertical axis wind shear effects for the WTE was discussed. The periodical effect introduces power fluctuations and mechanical stress on WT components. These effects depend on the number of blades and the rotation speed. The impacts of the horizontal and vertical shear on the torque and power fluctuations are of approximately 30% and 3%, respectively. From the analysis, it is concluded that these effects are essential for modeling the wind emulator.

3.2.10. $C_p-\lambda$ Approximation

The implementation of a realistic WT system was studied in [62]. It has been developed considering the electrical generator of a small wind turbine and the $C_p-\lambda$ curve. This work proposes an estimation of the $C_p-\lambda$ curve for a small wind turbine. This method avoids a wind tunnel, which is cost-effective but requires sufficient data with different wind conditions to obtain a $C_p-\lambda$ curve. The control algorithm for the wind emulator was developed in TMS320F28377. On the basis of the mechanical model, the $C_p-\lambda$ curve was obtained and the DSC was used to calculate the speed reference for an induction motor. The wind power and $\lambda$ are calculated from the mathematical model of a WT, and the value of $C_p$ is obtained from a lookup table. From this analysis, the power captured by the WT is computed at the same time, and the aerodynamic torque is obtained by dividing the captured power by the turbine speed. Finally, the turbine speed reference signal is fed through a variable-frequency drive to obtain the characteristics of the WT. This method is only valid for small wind turbines provided with the known $C_p-\lambda$ curve.

3.2.11. Indirect Field-Oriented Control

The WTE is designed based on an indirect field-oriented (IFO) induction motor drive [63]. The IFO drive has two control modes: speed and torque mode. The speed control mode produces the torque current command, which is derived from the difference between the reference rotor speed and the actual rotor speed. During torque control mode, the torque–speed and power–speed characteristics of the WT are designed by the known data provided by the vendor. In addition, the robustness of the torque control is enhanced by a robust weighting function. The torque control mode yields the current torque command from the deviation between the observed torque and command torque. Moreover, this IFO drive is employed for all types of generators. This scheme is only valid by known speed torque characteristics.

3.2.12. Fuzzy Based Field Oriented Control

A fuzzy-based FOC is proposed to enhance the dynamic behavior of the WTE [64]. The input and output variables of the fuzzy speed controller are the error, rate of change of error, and reference torque of the motor. In addition, the reference q axis and d axis stator currents are obtained through the fuzzy controller and reference rotor flux, respectively. Finally, these reference currents are passed through the current controller for the closed-loop operation of the wind turbine system.

3.2.13. Intelligent Relay Based Frequency Inverter

In [65], the WTE consists of a three-phase induction motor operated by a frequency converter. The converter controls the speed or torque of the induction motor by the direct torque control method. The shaft speed and torque are estimated with reasonable accuracy and sent to the intelligent relay. Finally, the mathematical model of the WTE computes the reference speed, and the intelligent relay sends this reference speed to the frequency converter.

3.2.14. Commercial Drive-Based Wind Emulator

In [66], the authors designed the WTE based on a squirrel cage induction motor as a prime mover and permanent magnet induction generator (PMIG) as electrical subsystems. The Danfoss drive is employed for controlling the prime mover. This emulator is compatible with different generators coupled to the prime mover, and it can test transient and steady-state dynamics for given wind speed and pitch angle. A Danfoss FC-51 microdrive is used to communicate the controlled resistor set, which is placed in between motor and drive. The Modbus remote terminal unit permits the drive to achieve the wind pattern from the computer. The motor torque or speed maintains the desired level by adequately assigning the gains in the PI controller.

A 22 kW induction motor-based wind emulator (downwind wind turbine) test rig was developed considering the tower shadow effect [67]. This test rig consists of a stochastic wind speed model, including the tower shadow effect (tubular and lattice shadow effect). In Gaia WT, the turbulence tower induces the wind speed deficit by 27%. For this reason, the study on the frequency response of an IM is essential. An induction motor is driven by variable speed Parker AC drive, and it can operate at both modes, such as speed and torque control mode. The Gaia WT is considered, and its works on fixed-speed topology. The torque control is adapted to a variable speed drive, and the dSPACE DS1103 controller is used to emulate the variable torque input due to the variable wind speed.

Table 2. Comparative analysis of different WTEs based on induction motor.

S. No	References	Variable Required to Build an Emulator (Motor)	Dynamics Consideration for Emulator					Converter and Control Technique for Emulator	Hardware Implementation	Emulator Accuracy
			${\mathit{C}}_{\mathit{p}}-\mathit{\lambda }$ Curve	Tower Shadow Effect	Wind Shear Effect	Torque Compensation	Inertia Compensation
1	[35,36]	Torque and speed	✓					IGBT inverter with PI controller	Intel 80C196KD microcontroller	*
2	[37]	Speed	✓	✓				Commercial frequency inverter (Danfoss-VLT 5022)	dSPACE	**
3	[38]	Torque and speed	✓					Inverter with direct torque control	dSPACE	*
4	[39]	Speed		✓	✓	✓		Inverter and dynamic torque control	DSP TMS320F2833	**
5	[40]	Speed					✓	Inverter and dynamic torque control	DSP dsPIC30f4011	*
6	[78]	Speed	✓					Inverter and torque control	dSPACE	*
7	[41]	Speed	✓	✓	✓			Inverter and torque control	–	**
8	[43]	Speed	✓					Torque-controlled drive	PLC	*
9	[44]	Speed and torque	✓					Frequency converter and direct torque control	PIC32MX360F512L	*
10	[45,46,47,48] ${}^a$	Speed				✓	✓	Off-the-shelf inverter	DSP	***
11	[49]	Speed				✓	✓	IGBT inverter with PI controller	DSP	*
12	[50]	Speed	✓	✓	✓			3-phase IGBT inverter	DSP	***
13	[51] ${}^c$	Speed	✓			✓		Frequency converter	PLC, DSP, and SCADA	***
14	[52]	Speed and torque	✓					3-phase inverter	DAQ NI 6211	*
15	[53,54]	Torque	✓					Frequency inverter	DSP TMSF28335	*
16	[55]	Torque	✓					Torque control	DSP TMSF28335	*
17	[56]	Speed				✓	✓	AC drive	DSP	***
18	[57]	Speed					✓	VACON variable-frequency drive	DCS based on Beckhoff PLC	**
19	[58]	Speed	✓					3-phase IGBT Inverter with Volt/Hertz control	dSPACE	*
20	[59]	Speed	✓					AC drive	PIC16F876A	*
21	[60]	Torque and Speed			✓			Variable speed drive	LabVIEW	***
22	[61] ${}^d$	Torque and Speed				✓		Variable speed drive	LabVIEW	****
23	[62]	Torque and Speed	✓					Variable speed drive	DSP TMS320F28377	*
24	[63]	Torque and Speed	✓					Inverter	DSP	*
25	[64]	Speed	✓					Field-oriented control	–	*
26	[65]	Speed	✓					Frequency converter	SCADA	**
27	[66]	Speed	✓					IGBT driver board	Danfoss FC-51 Microdrive	**
28	[67]	Speed		✓				Variable-speed AC drive (Parker SSD Drive)	dSAPCE	***

^a Coefficient of viscosity of WT and IM is included. ^c Inertia and loss coefficient is included. ^d 24 DOFs and rotor teetering effect included. * Less, ** Moderate, *** High, and **** Very high.

shows the different WTEs based on IM.

From Table 1 and Table 2, most of the designed WTEs utilized the $C_p-\lambda$ curve approximation, which enables the easy implementation of the characteristics of the emulators. However, the actual $C_p-\lambda$ curve is highly nonlinear; at the same time, the accuracy of these emulators is inadequate. Some authors have utilized the inertia and torque compensation schemes [13,18,19,44,45,46,47,48,56] which are more appropriate to capture the characteristics of the emulators. Furthermore, some studies have included the wind and tower shadow effect [12,39,50,78] in the design of emulators. This approach could be more accurate for the emulators to capture the dynamics of the actual WT. In addition, references [60,61] used the FAST simulator to design the WTEs, and these schemes included several DOFs (24 DOFs), which are used to analyze the WTEs from all perspectives.

3.3. Doubly Fed Induction Motor

The authors in [68] developed the WTE based on the doubly fed induction motor (DFIM). A stator flux-oriented method was applied to control the rotor of DFIM [68] and presented in Figure 10. The model reference adaptive system (MRAS) technique estimates the rotor speed using the rotor current. The reference torque is obtained from the wind turbine characteristics. Thus, the rotor of DFIM is adequately controlled to emulate the required shaft torque.

3.4. Servo Motor

The wind emulator was developed by connecting the two servo motors in back-to-back configuration [69]. A power electronics converter controls each servo motor. The effect of wind turbulence and the mechanical characteristics of the drive train are modeled in the FPGA-based control board. The reference torque is generated from the model and sent to the torque control of the servo motor. Gokhale et al., developed a wind turbine mathematical model in RSCAD software [70]. The wind speed and rotor speed are utilized to obtain a reference torque. An encoder measures the rotor speed through a digital card of real time digital simulator (RTDS). The reference torque generated in the RSCAD model is converted into an appropriate voltage signal and sent to the servo drive through an analog card of RTDS. The servo drive is set to torque control mode and follows the torque control reference.

3.5. Permanent Magnet Synchronous Motor

It is a well-known fact that the inertia of an actual wind turbine is much greater than a WTS. Therefore, in [71], an inertia compensation-based WTS was designed for a permanent magnet synchronous motor. Equations (23) and (24) refer to the static and dynamic behavior of the actual wind turbine and WTS, respectively. First, the real reference mechanical torque is obtained by setting the speed of the actual WT and WTS to $\omega$. Then, in Equation (25), the second term becomes zero during the steady-state operation. Finally, the WTS performs the actual WT by controlling the reference mechanical torque in Equation (25). Furthermore, a low pass filter is added to the transient torque compensation to attenuate the high-frequency noise.

$$\frac{T_r}{n}=\left(\frac{J_r}{n^2}+J_g\right)\frac{d{\omega }_{g1}}{dt}+T_e$$

(23)

$$T_m=\left(\frac{J_m}{J_g}\right)\frac{d{\omega }_{g2}}{dt}+T_e$$

(24)

$$T_m=\frac{T_r}{n}-\left(\frac{J_r}{n^2}-J_m\right)\frac{d\omega }{dt}$$

(25)

where $T_r$ is the aerodynamic torque, $T_c$ is the compensation torque, $T_m$ is the mechanical torque, and $T_e$ is the electromagnetic torque In [72], the delay due to the speed control loop is eliminated by improving the dynamic compensation loop. In [73], the wind emulator utilizes surface-mounted PMSM (SPMSM) with a control drive. An encoder is employed to find the position of the rotor, which it sends to the speed estimator to estimate the rotor speed. Then, the relationship between the reference torque, wind speed, and rotor speed is derived, and it helps to obtain the reference torque. Finally, the torque control or speed control drives the SPMSM as a WTE.

4. Computational Based Wind Turbine Emulator

This section presents an implementation of the WTE as software on computational devices. The DFIG-based wind emulator is developed on the FPGA platform, and the dynamic response is derived by the fourth-order Runge–Kutta method [74]. In addition, the complete system model was developed in VHDL and synthesized by Xilinx ISE. Furthermore, an unrealistic inertia value was considered to minimize the simulation time. Finally, the same DFIG model was implemented in the Simulink platform, and this analysis concludes that the FPGA has a 40× simulation speed than PC-based simulation.

In [75], the authors presented the implementation of the WTE in a real-time simulation using RT-LAB with three CPUs. The WTE system was divided into three parts and implemented on different CPUs. Finally, all CPUs were conducted synchronously and run at 2.5 GHz with a simulation step of 10 $\mathsf{\mu }$s. Compared with conventional simulation, the RT-LAB has the advantage of accelerating the execution and increasing simulation efficiency.

A LABVIEW DSC module-based real-time simulation of wind emulators was presented in [76]. In addition, a control scheme for the emulator is designed in the MATLAB/Simulink platform, and a simulation toolkit was used to communicate between the power and control module.

A flexible and adequate WTE is essential to implement a model-based controller design. In [77], the authors propose the real-time heterogeneous WTE, which combines a discrete time step simulation and linear parameter varying (LPV) model. In addition, a 5 MW NREL model was considered and implemented on an NVIDIA Jetson TK1 board. This LPV model has 16 DOFs, including the dynamics of blades, drive train, yaw, and tower modes. Furthermore, a set of locally linearized models were considered for the different regions of WT, and it is utilized to derive the LPV model. The real-time discrete time (RTDT) simulation is divided into four-time frames: input transfer, heterogeneous LPV (H-LPV) engine, interrupt time, and output transfer. Moreover, the H-LPV engine plays a vital role in the heterogeneous WTE in the subroutines to maximize the execution performances and manage to schedule the vector arithmetic operations. Finally, the heterogeneous WTE has a 2.73× acceleration speed compared to MATLAB, and the empirical results are similar to real-world scenarios.

Table 3. Comparison of computational WTEs.

S. No.	Reference	Emulator Model	Simulation Software	Platform	Computational Speed	Accuracy of the Emulator
1	[74]	DFIG-based emulator	VHDL and synthesized by Xilinx ISE	FPGA	40-fold faster simulation speed than the PC-based simulation	**
2	[75]	BLDC-based emulator	Matlab/Simulink	RT-LAB	-	**
3	[76]	Real turbine model with gear box	LabVIEW DSC module and Matlab/Simulink	LabVIEW simulation interface toolkit	-	**
4	[77]	5 MW NREL	Matlab/Simulink	NVIDIA Jetson TK1 board	2.73-fold faster speed compared to MATLAB	****

** Moderate, and **** Very high.

presents the comparison of computational WTEs.

Table 1 and Table 2 show the comparative analysis of the different WTEs available in the literature. These tables show the physical significance of the parameters required to design WTEs. The majority of emulators are developed on the basis of the power curves of the real WT. However, these power curves are highly nonlinear and emulators are developed by approximating the power curves. In addition, most emulators are designed by coupling the motor with generators to emulate the characteristics of real WT. The inertia of the real WT is much higher than that of the WTE. As such, it is essential to compensate for the inertia of the machine. Therefore, the authors considered both inertia and torque compensation to develop the emulator. Some research has been conducted considering all the real physical parameters such as a tower shadow and wind shear effects. Table 3 presents the computation-based WTE. These emulators are developed on the computational platform with simulation software. This method is cost-effective compared to other methods. Moreover, the user can implement a different degree of freedom, which can include the dynamics of the blades, the drive train, and the tower dynamics. In this method, the electrical dynamics with control schemes are not included in the development of WTE.

5. Comparative Study

Numerous studies have shown that the development of WTEs replicates the characteristics of wind energy conversion systems. Consequently, a detailed comparative analysis has been presented based on the following criteria, such as (1) realization cost; (2) accuracy; (3) level of complexity; and (4) hardware implementation. Furthermore, these discussions support that WT researchers intensify the dynamic and steady-state performance of WTEs.

5.1. Realization Cost

Implementation cost is the vital parameter to be assessed while developing the WTE. The design of WTE involves different motors such as the DC motor, induction motor, servo motor, doubly fed induction motor, and brushless DC motor. A DC motor is relatively large, more expensive, and demands long-term maintenance. Nonetheless, the control of the DC motor can be simply matched with the characteristics of the actual wind turbines. On the other hand, IMs are less expensive, robust, and have a low maintenance cost. However, a complex control system is required to deal with the inherent nonlinearities. The real-time controllers for the wind emulator are implemented by dSPACE [12,32,38,67], PLC [43,51,57], and Danfoss FC-51 [66]. Unfortunately, these controllers necessitated additional costs along with the motor.

5.2. Accuracy of the Emulator

The WTE aimed to replicate the steady-state and dynamic characteristics of the wind turbine. These emulators are employed to test and analyze the control strategy, power converter topologies, and integration of multiple renewable energy sources. Therefore, the effectiveness of the WTEs was investigated according to the accuracy of the emulation of static and dynamic characteristics rather than the precision of the WTEs. Various methods are presented in the literature to more accurately design the wind emulator. Among the methods, direct torque control, open loop control, and polynomial approximation showed insufficient accuracy and high irregularity in the characteristics of the WTs. On the other hand, the torque and inertia compensation methods provide better wind turbine characteristics, which are very similar to the actual wind turbine.

5.3. Level of Complexity

The level of complexity in WTEs is a significant factor and is utilized to find the accuracy of the emulator. As such, to examine the emulators, the following criteria are employed: (1) parameters required for the implementation of the WT model; (2) parameters required for the implementation of generator dynamics; and (3) control structure and power converter topology. An accurate emulator is designed with several dynamics. For example, the works in [9,10,11,35,36,38] included WT dynamics such as wind speed, turbine power, and blade pitch angle. In these cases, the complexity level is relatively low with a high modeling error. However, the references [13,14,15,16,17,45,46,50,56] have dynamics such as wind shear, tower shadow, torque compensation, and inertia compensation. Including the wind shear and tower shadow effect introduces a high level of complexity with a high modeling accuracy. Furthermore, the controller implementation for a real-time wind emulator was coded on different platforms. Therefore, the high accuracy of the emulator comes with the cost of high complexity.

5.4. Hardware Implementation

The hardware implementation of the wind emulator is a coupling of two machines, such as a motor (DC motor or Induction motor) and a generator (PMSG or DFIG). The motor is considered a prime mover for executing the WT dynamics. Several parameters are involved in designing a wind emulator, and the accommodation of those parameters for hardware implementation is quite challenging. For example, the DC motor-based wind emulator utilizes a simple chopper circuit or buck converter, and the hardware of these approaches is implemented by microcontrollers [14,19]. In [10,11,32,37,38,58], a dSPACE-based real-time emulator was utilized, and a DSP processor-based wind emulator design was incorporated in [14,17,18,39,42,49,50], LabVIEW [26,60,61], FPGA [22], and Arduino [30,31]. Similarly, in [66], the Danfoss FC-51 microdrive was adapted to realize a wind emulator.

Table 4. Comparison of different methodologies of WTEs.

S. No.	Methods	References	Cost	Complexity	Hardware Implementation	Accuracy
1	Current/power reference	[7,9,10,20,35,36]	Less	Less	Easy	Moderate
2	Open loop	[21,22,37]	High	Moderate	Easy	Moderate
3	Observer based	[19,38]	Less	Moderate	Moderate	Less
4	Torque compensation-based torque control	[13,14,15,16,17,18,39,40,41,42,78]	Moderate	Less	Easy	High
5	Improved inertia compensation	[56]	Less	Moderate	Easy	High
6	Lookup table	[58]	Less	Less	Easy	Less
7	Power and speed relation	[59]	High	Less	Easy	Less
8	$C_p-\lambda$ approximation (polynomial approximation)	[23,24,62]	High	Less	Easy	Less
9	FAST Simulator	[60,61]	Less	Less	Easy	High

summarizes the different methodologies utilized for developing WTE based on the following parameters: cost, complexity, hardware implementation, and accuracy. The efficacy of each method has been quantified in terms of less, moderate, and high. From this table, the torque compensation, inertia compensation, and FAST simulator-based WTEs are highly efficient in emulating the characteristics of WT, which is almost similar to PWT. In addition, the open-loop scheme is also suitable for WTE but requires parameter adjustment for every wind variation. The other methods, such as current/power reference, $C_p$ approximation, and lookup table, are easy to implement, but these methods are highly inaccurate.

The WTEs are developed in the laboratory environment. There are two categories of emulators available in the literature. Most of the developed WTEs fall into the category of a combination of physical and computational emulators. Therefore, the dynamic characteristics of the actual WT were implemented in the motor, which is coupled with a generator. This kind of set-up is widely used because it can test and analyze the control schemes and power converter topologies developed by researchers. Furthermore, some authors developed WTE based on the pure computational method. The entire emulator is developed in computational software and can adapt the various DOFs from the actual WT. This method is cost-effective and the speed of the emulator depends on the computation platform and simulation software. However, the practical significance and accuracy of this method are relatively lower than the combination of physical and computational emulators.

6. Conclusions and Future Recommendations

This paper reviews the various methodologies adapted to realizing the WTE. In addition, a detailed investigation was carried out by analyzing the different WTEs and their ability to reproduce the static and dynamic characteristics of the PWT. All methodologies were compared on the basis of cost, accuracy, complexity, and hardware implementation. The following points are summarized from the analysis of the various studies available in the literature:

• The majority of real-time WTEs focus on electrical systems, whereas WT mechanical systems either model steady-state or low-order models. These models ignore the rotor blades, tower structure, and drive train dynamics.
• Some authors have linearized the dynamics of the mechanical system at a single point, and the power curves are incorporated to derive the torque. However, these assumptions are no longer valid because the WTs operating curves are highly nonlinear.
• Most of the emulators available in the literature come under a combination of physical and computing-based emulators. Thus, these emulators are lab-scale and case-specific.

The following recommendations could be considered for further development in the WTE.

• A sufficient number of WT mechanical dynamics should be considered for modeling the WT mechanical subsystem so that the nonlinear nature of the PWT could be addressed in the emulator.
• As discussed earlier, in most WTEs, the electric motor acts as a prime mover. Therefore, the emulator requires inertia compensation due to the lower inertia of the electric motor.
• An emulator can shift the operating point corresponding to changes in actual wind speed and the capability of adapting the new control strategies and power converter topologies for further development in wind power systems.
• An emulator could be capable of incorporating the structural dynamics of WT without local linearization.
• The transient modes of the WT mechanical system have to be considered for developing WTEs. In addition, future research could focus on hybrid wind emulators.