Optimal Control and Operation Strategy for Wind Turbines Contributing to Grid Primary Frequency Regulation

Transcription

1 applied sciences Article Optimal Control and Operation Strategy for Wind Turbines Contributing to Grid Primary Frequency Regulation Mun-Kyeom Kim School Energy System Engineering, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul , Korea; Tel.: Received: 23 June 2017; Accepted: 5 September 2017; Published: 8 September 2017 Abstract: This study introduces a frequency regulation strategy to enable participation wind turbines with permanent magnet synchronous generators (PMSGs). The optimal strategy focuses on developing frequency support capability PMSGs connected to power system. Active power control is performed using maximum power point tracking (MPPT) and de-loaded control to supply required power reserve following a disturbance. A kinetic energy (KE) reserve control is developed to enhance frequency regulation capability wind turbines. The coordination with de-loaded control prevents instability in PMSG wind system due to excessive KE discharge. A KE optimization method that maximizes sum KE reserves at wind farms is also adopted to determine de-loaded power reference for each PMSG wind turbine using particle swarm optimization (PSO) algorithm. To validate effectiveness proposed optimal control and operation strategy, three different case studies are conducted using PSCAD/EMTDC simulation tool. The results demonstrate that optimal strategy enhances frequency support contribution from PMSG wind turbines. Keywords: de-loaded control; frequency regulation; kinetic energy reserve control; maximum power point tracking; particle swarm optimization; permanent magnet synchronous generator 1. Introduction Over past few years, wind energy has received significant attention as one most promising renewable energy sources given adverse environmental impacts conventional energy sources [1,2]. Both permanent magnet synchronous generator (PMSG) wind turbines and doubly fed induction generator (DFIG) wind turbines have been widely used. PMSG wind turbines are advantageous over DFIG wind turbines due to ir highly reliable operation, lower maintenance expenses, and smaller weight with a simpler structure. Moreover, PMSG can connect to turbine without use a gearbox [3,4]. Therefore, many recent studies have focused on PMSG wind turbines. In power systems with high wind power penetration, frequency control is a critical issue from viewpoint stable system operation. The wind power system is required to provide a reliable frequency response to support grid and reduce costs associated with reserve power. Variable-speed wind turbines are commonly operated to achieve optimal efficiency for maximum power output using maximum power point tracking (MPPT) control technology [5]. As a result, wind turbines do not have an inherent frequency regulation capability to supply additional reserves [6]. A sufficient reserve margin is an important requirement to participate in frequency regulation, which is absent in a wind turbine generator. Many researchers have investigated frequency support capability wind turbine generators to supply power margin. To provide an inertial response, authors [7 9] proposed an additional Appl. Sci. 2017, 7, 927; doi: /app

2 Appl. Sci. 2017, 7, power control loop for variable-speed wind turbines. Furr, research in [10,11] has helped develop a concept that directs a portion rotational energy and blade to short-term active power support, which could aid in reducing system frequency drop after a generation deficit. Anor option is to utilize a de-loaded control operation to provide primary frequency response. In [12], authors suggested an operation at de-loaded power curve a wind turbine to supply power margin. Various methods are available for de-loaded operations; for example, a combination pitch angle control and rotational speed control [13 17]. However, increasing dynamic characteristics could affect life pitch angle controller and increase system cost. The authors [18 20] considered a fixed percentage wind turbine de-loading. De-loaded control decreases power output wind turbine generator but increases primary frequency support capability during frequency disturbances. However, de-loading is restricted to between 10% and 20%, depending on available wind speed and rotor speed limit constraints [18]. In [21,22], authors addressed kinetic energy (KE) discharge control that incorporated a wind farm to supply additional reserves for frequency support. Two options have been reported in previous studies for participation wind turbines during a frequency event: de-loaded control and inertial response for primary frequency control. The inertial control utilizes KE in rotating mass wind turbine to provide a temporary frequency response. The de-loaded control allows wind turbine to run in de-loaded operation mode to reserve partial wind power for primary and secondary frequency regulations. However, an unsuitable margin power would rapidly decrease rotational speed wind turbine. In addition, improper control energy regained during acceleration may cause anor frequency disturbance [23]. Overall, main focus previous studies has been on frequency support capabilities wind turbine generators. Very few published papers or documents emphasize comprehensive frequency regulation capacity PMSG based on rotor speed-based control under different wind speeds or enhanced inertial responses based on active power-based control when operated in MPPT mode, especially for potential impact frequency regulation on mechanical components wind turbines. Therefore, this research addresses major issues with goal maximizing PMSG-WTG performance for frequency regulations in accordance with its specific operation characteristics and control structure. Concurrently, it is necessary to research a frequency regulation strategy to effectively maximize stored KE. The authors [24,25] suggested an operational strategy for maximizing amount KE during a frequency disturbance. However, se studies have not described optimization techniques in detail. This study proposes a new optimal control and operation strategy for a PMSG wind turbine to support system frequency in wind energy production. The machine-side converter (MSC) and grid-side converter (GSC) are designed to control DC link voltage using vector control, and obtain maximum power output from PMSG wind turbines. The active power control operates turbine with MPPT and achieves power margin through de-loaded control. Once se are achieved, KE reserve control establishes frequency regulation to provide stable PMSG wind turbines. To increase contribution wind turbines to support frequency stability via a temporary release energy, KE optimization method performs a KE maximization for various wind speeds using particle swarm optimization (PSO) algorithm. The operation PMSG wind turbines is n optimized to extract more KE, which can be released into power system in case a frequency event. The optimization strategy assumes that wind farm system performs continuous output control with monitoring and control systems for individual PMSG wind turbines. With wind speed, this strategy calculates maximum possible output, optimal rotational speed, and KE reserve in each wind turbine. The de-loaded power reference from an individual PMSG wind turbine that maximizes KE reserve wind farm is obtained by satisfying output command system operator. The frequency support capability PMSG system with proposed KE reserve control strongly depends on KE reserve obtained from de-loaded operation and generating margin. Therefore, frequency response a system can be improved after organizing de-loaded operations

3 Appl. Sci. 2017, 7, individual PMSG systems with regard to KE reserve when wind farm is de-loaded. The KE optimization strategy, which maximizes sum KE reserves at a wind farm, is adopted to determine de-loaded power reference for each PMSG wind turbine. The rest this work is structured as follows. Section 2 presents system modeling and converter control. Section 3 presents proposed control and operation strategy to support frequency regulation using PSO. Section 4 describes a comparison simulation results or cases to assess performance proposed strategy. Section 5 presents concluding remarks. 2. System Modeling and Converter Control 2.1. Wind Turbines and Modeling PMSG The operational performance a wind turbine can be modelled from mamatical relationship between wind speed and mechanical power as follows: P m = 0.5ρπR 2 S 3 wc p (λ, β) (1) λ = ω mr S w (2) Here, power coefficient C p represents rotor aerodynamics as a function both pitch angle rotor blades and tip speed ratio (TSR). The PMSG modeling facilitates generation electricity from mechanical energy and a dynamic model is derived using a two-phase synchronous reference frame, in which q-axis is 90 ahead d-axis in accordance with direction rotation. In three-phase systems, such as PMSGs, each phase quantity has a time-varying characteristic. Using Park s transformation, which is projection phase quantities onto a rotating two axes reference frame, AC quantities can be converted to DC quantities with time independence. The abc to dq0 transformation is represented in matrix form: u d u q u 0 = 2 3 cos(θ r ) cos ( θ r 2π ) 3 sin(θ r ) sin ( θ r 2π ) cos ( θ r + 2π ) 3 sin ( θ r + 2π ) 3 u a u b u c (3) The inverse Park s transformation is: u a u b u c = 2 3 cos(θ r ) sin(θ r ) cos ( θ r 2π ) 3 sin ( θ r 2π 3 cos ( θ r + 2π ) 3 sin ( θ r + 2π 3 2 ) 2 2 ) u d u q u 0 (4) In Equations (3) and (4), u abc and u dq0 express stator voltages, stator currents or flux linkages AC machines, respectively. Applying equality u 0 = 0 for balanced conditions, voltage function PMSG in d q axis reference frame [4] is given by V ds = R s I ds + L d di ds dt V qs = R s I qs L q di qs dt + ω e L q I qs (5) ω e L d I ds + ω e Ψ f (6) To complete PMSG mamatical model, its mechanical equation is required [26]; electromagnetic torque equation is described by T e = 1.5p n [ (Ld L q ) Ids I qs + I qs Ψ f ] (7)

4 Appl. Appl. Sci. Sci. 2017, 2017, 7, 7, To complete PMSG mamatical model, its mechanical equation is required [26]; 2.2. MSC and GSC Control electromagnetic torque equation is described by Figure 1 shows block diagram MSC and GSC. As shown in Figure 1a, MSC control mainly involves adjustment PMSG Tetorque = 1.5p nand ( Ld rotational Lq ) IdsIqs speed IqsΨ + t set special operating conditions. (7) The MSC can be adjusted to regulate PMSG speed at a reference value due to MPPT. The d q axis current component indicates components torque and flux, and stator current reference 2.2. MSC and GSC Control value (zero) in d-axis is assigned to achieve maximum torque and maximize resistive losses. The required Figure 1 d q shows components block diagram voltage MSC vector and aregsc. derived As from shown twoin proportional Figure 1a, MSC integral control (PI) current mainly involves controllers, which adjustment are applied PMSG to control torque and speed rotational generator speed set and turbine. special operating After PI conditions. current controllers, The MSC can compensation be adjusted components to regulate are PMSG added speed to improve at a reference transient value response, due to as shown MPPT. in The Figure d q axis 1. Components current component 1 and indicates 2 are ω e L q I qs components and ω e L d I ds + torque ω e Ψ f, and respectively. flux, and They stator are obtained current reference from Equations value (zero) (5) in and (6). d-axis Theis Park assigned andto Clarke achieve transformations maximum torque areand applied maximize to d- and resistive q-components losses. The to generate required d q three-phase components current references voltage forvector MSC. are The derived speedfrom reference, two ωproportional m = λ S w /R integral calculated (PI) current from Equation controllers, (2), is which restricted are applied to maximum to control angular speed speed (3.5 generator rad/s). Here, and turbine. we use After pulse width PI modulation current controllers, (PWM) to compensation produce a switching components signal are foradded MSC control. to improve transient As shown response, in Figure as shown 1b, GSC in Figure control 1. maintains Components a constant 1 and 2 DCare link ω elvoltage qiqs and regardless ω eld Ids +ωe Ψ f, magnitude respectively. They PMSG are power obtained and from generates Equations any(5) reactive and (6). power The Park to obtain and Clarke a power transformations factor unity. are To applied develop to control d- and mode q-components for GSC, ato vector generate control method three-phase is applied current withreferences a for frame MSC. grid The voltage, speed reference, utilizing independent ω m = λ S w R control calculated from active Equation power and (2), reactive is restricted powerto between maximum GSC and angular grid. speed (3.5 rad/s). Here, we use pulse width modulation (PWM) to produce a switching signal The for voltage MSC control. equations for dynamic model grid connection in d q reference frame [27] can be As expressed shown in asfigure 1b, GSC control maintains a constant DC link voltage regardless di magnitude PMSG power and dg V dg generates = V di Rany g I dg reactive L g power + ω g Lto g I qg obtain a power factor unity. (8) To develop control mode for GSC, a vector control dt method is applied with a reference frame grid voltage, utilizing independent di qg V qg = V control qi R g I qg L active g power ω g L g I and reactive power between dt dg (9) GSC and grid. V dc ω m I ds I qs ω m,ref I ds,ref = 0 I qs,ref V qs V ds (a) Figure 1. Cont.

5 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, V dc,ref I dg,ref V dc V di V qi V dg V qg I dg I qg, ref I qg V dc (b) Figure 1. Block diagram machine-side converter (MSC) and grid-side converter (GSC). (a) MSC Figure 1. Block diagram machine-side converter (MSC) and grid-side converter (GSC). (a) MSC control system; (b) GSC control system. control system; (b) GSC control system. The voltage equations for dynamic model grid connection in d q reference frame The [27] active can be and expressed reactive powers as [4] are respectively expressed as follows: Vdg di dg Vdi Rg I dg Lg + dt P g = 1.5V dg I dg (10) = ω (8) di qg Vqi Rg I qg Lg dt g Lg I qg Q g = 1.5V dg I qg (11) The q-axis current component Vqg= GSC is used to control ω g Lgflow I dg reactive power, and(9) d-axis current component GSC is used to regulate DC link voltage using a PI controller. In this study, The active grid q-axis and reactive current is powers set to zero [4] are to prevent respectively generation expressed any as reactive follows: power. A phase-locked loop (PLL) circuit is applied to synchronize Pg = three-phase 1. 5V dg I voltage with grid voltage. To transfer dg (10) all active power obtained from wind turbine to grid, DC link voltage should remain constant, as shown in following constraint Qg [5]. = 1. 5V dg I qg (11) The q-axis current component GSC is used to control flow reactive power, and d-axis current component GSC C dv dc is = used P t to P g regulate DC link voltage using a (12) dt V PI dc V dc controller. In this study, grid q-axis current is set to zero to prevent generation any reactive power. If A phase-locked wind turbineloop power (PLL) P t circuit and is grid applied power to synchronize P g are equal, three-phase DC link voltage must with be constant. grid voltage. Compensation To transfer all components active guarantee power obtained stable from and decoupled wind turbine active and to reactive grid, power DC control. link voltage Components should remain 1 andconstant, 2 are ω g Las g I qg shown and ω in g L g following I dg, respectively. constraint These [5]. terms can be added to enhance transient response power system (Equations (8) and (9)). dv P dc Pt g C = (12) dt V V If wind turbine power P t and grid power P g are equal, DC link voltage must be constant. Compensation components guarantee stable and decoupled active and reactive power control. Components 1 and 2 are ω glgiqg and ωg Lg Idg, respectively. These terms can be added to enhance transient response power system (Equations (8) and (9)). dc dc

6 Appl. Sci. 2017, 7, Proposed Optimal Strategy 3.1. Control Mode Wind Turbine Active Power Control Maximum power point tracking (MPPT) control is usually employed to maximize energy conversion efficiency a wind turbine by regulating rotational speed a variable-speed wind turbine. Depending on wind aerodynamic conditions, an optimal operating point exists that may allow extraction maximum power from turbine. This point is called maximum power point (MPP). The power captured with wind turbine can be substantially maximized by adjusting coefficient C p, which represents aerodynamic efficiency wind turbine and is dependent on TSR λ. The wind turbine can generate maximum power as expressed in Equation (1). Then, maximum mechanical output power turbine is given by P max = 0.5ρπR 2 S 3 wc p,opt ( β, λopt ) The optimal TSR mechanical output power can be represented as follows: (13) λ opt = ω m,optr S 3 w (14) For PMSG wind turbine, maximum value power coefficient (C p, opt = ) is obtained as an optimal value TSR (λ opt = 6.9), calculated using Equation (14). The coordination TSR to its optimal value with power coefficient reaching C p, opt guarantees maximum power extraction. Equation (15) provides formulation for maximum power determined using MPPT control: P max = ω 3 m,opt K opt (15) with K opt = 0.5ρπ ( Cp,opt λ 3 opt ) R 5 (16) This MPPT control allows wind turbine to extract as much power as possible for each wind speed. However, when power system begins to impose requirements on wind power for objectives, such as frequency regulation, or operational modes are required. For wind turbine generators to improve in frequency support capability, y must secure sufficient reserves. In this study, PMSG wind turbines are operated at de-loaded power point (over speeding) instead at MPP. The de-loaded control mode can also be implemented using a combination control techniques for rotational speed and pitch angle. Figure 2 shows MPPT and de-loaded power extraction curves for a PMSG wind turbine at various wind speeds. In setting maximum de-loaded values for a PMSG wind turbine, both current rating converters and maximum allowable limit rotational speed need to be considered. In 5% de-loaded control for a PMSG wind turbine, power output is P del = 0.95P max. To provide a reliable operation range PMSG, this study focuses on maximum rotational speed 1.2 p.u., with a reduction in certain fatigue loads as an additional objective. Figure 3 shows de-loaded controller. To utilize proposed de-loaded control mode, power margin P margin a wind turbine is represented as follows: P margin = P max P del (17)

7 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, 7, Figure Maximum power power point point tracking tracking (MPPT) (MPPT) and de-loaded and de-loaded power power extraction extraction curves for curves different for wind different speeds. wind speeds. Figure 3. De-loaded controller KE Reserve Control To provide additional active active power power to to grid, grid, KE reserve KE reserve control control utilizes utilizes energy stored energy in stored rotating masses rotating masses generator. generator. This concept This has concept advantage has advantage being able to being make able a relatively to make large a relatively contribution large contribution to short-term to frequency short-term control. frequency However, control. excessive However, discharge excessive KE discharge with aim KE with making aim a large making contribution a large continuously contribution during continuously participation during participation wind turbine in frequency wind regulation turbine in not frequency only decreases regulation rotational not only speed decreases wind rotational turbine significantly, speed but wind it may turbine also make significantly, wind but turbine it it may unstable. also make Therefore, wind for stable turbine operation unstable. Therefore, a PMSG wind for generator, stable operation it is crucial a to PMSG appropriately wind generator, set output it it is is crucial provided to appropriately to system set during output frequency provided control. to In this system study, during KE reserve frequency control control. for frequency In this study, support KE is used reserve along control with for frequency de-loaded support controlis to used supply along with amount de-loaded power reserve control for to supply primary amount frequency power response. reserve for primary frequency response. Figure 4 shows a diagram proposed KE reserve control method. de-loaded control makes rotational speed increased with with to to optimal optimal rotational rotational speed speed for a for given a given wind speed, wind speed, and it means and it it that means that KE stored KE in rotating stored in masses rotating is also masses increased is is also when increased PMSG system when PMSG is de-loaded. system Figure is is de-loaded. 5 shows Figure KE reserve 5 shows controller. KE If reserve frequency controller. decreases If If because frequency frequency decreases disturbance because in frequency power system, disturbance de-loaded in control power mode system, switches de-loaded to KE control reserve mode mode switches using KE to reserve KE reserve control. mode This using change KE immediately reserve control. increases This change output immediately P increases output Pdel Pdel PMSG as del PMSG power increment P KE, and maintains power its output increment by participating PKE, and maintains frequency its output regulation by participating during a support in frequency time T support regulation. In or during words, a support P out is time Tsupport. In or words, Pout Pout is is increased by increased by P KE and rotational speed decreases PKE and because rotational KE reserve speed control. decreases because KE reserve Because control. rotational speed at de-loaded power point is higher than optimal rotational speed, de-loaded control can significantly increase KE reserve, which is given by ( ) KE res = 0.5J ωdel 2 ω2 m,opt (18)

8 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, 927 Appl. Sci. 2017, 7, Figure 4. KE reserve control scheme. 4. Figure 4. KE reserve control scheme. Figure 5. KE reserve controller. Because rotational speed at de-loaded power point is higher than optimal rotational speed, de-loaded control can significantly increase KE reserve, which is given by ( ) Figure 5. KE reserve 2 controller. 2 Figure KE res5. = KE 0.5reserve J ωdel controller. ωm, opt (18) Because rotational speed at de-loaded power point is higher than optimal rotational Because Pitch Angle rotational Control speed at de-loaded power point is higher than optimal rotational speed, speed, Pitch de-loaded de-loaded Angle Control control control can can significantly significantly increase increase KE KE reserve, reserve, which which is is given given by by When wind speed is above rated speed, 2 pitch 2 angle control mode adjusts pitch When wind speed is above KErated res speed,.5j ( ωdel 2 ωpitch m2, opt ) angle control mode adjusts (18) pitch angle reby reducing power obtained KE res from = 0.5J ( ωdel wind. ωmfigure, opt ) shows pitch angle controller, (18) angle reby reducing power obtained from wind. Figure 6 shows pitch angle controller, which is based on PI control for processing error between rate value and angular speed which is based on a PI control for processing error between rate value and angular speed generator/turbine. Pitch Angle Control When wind turbine operates with power optimization strategy or power generator/turbine. Pitch Angle Control When wind turbine operates with a power optimization strategy or power limitation, this control mode limits rotational speed to maintain reliable operation range for limitation, When this wind control speed mode is limits above rotational rated speed, speed to pitch maintain When wind speed is above rated speed, pitch angle a angle control reliable control mode operation mode adjusts range adjusts for pitch PMSG. This work focuses on maximum rotational speed 1.2 p.u. with reduction certain pitch angle PMSG. angle reby This work reby reducing focuses reducing on power power obtained maximum obtained from rotational from wind. speed wind. Figure 1.2 Figure 6 shows p.u. with shows pitch a reduction pitch angle angle controller, certain fatigue loads as an additional objective. This controller adjusts pitch angle based on comparison controller, which fatigue which is loads is based as based on an on additional a PI PI control objective. control for for processing This controller processing error adjusts error between between pitch rate angle rate value based value and and angular on a comparison angular speed speed ω m. The reference pitch angle β generator/turbine. generator/turbine. When When wind ref is determined by PI controller. The operating rotational speed ωm. The reference pitch angle βref wind turbine is determined turbine operates by operates with with PI a power controller. power optimization The operating optimization strategy rotational strategy or or power must be limited between 0.7 and 1.2 p.u. The pitch angle control a PMSG wind turbine can power be limitation, speed must limitation, this be this control limited control mode between mode limits 0.7 limits and rotational 1.2 p.u. The rotational speed pitch speed to to maintain angle control maintain a reliable a PMSG reliable operation wind operation range turbine range for for can summarized as follows: PMSG. be summarized PMSG. This This work as work focuses follows: focuses on on maximum maximum rotational rotational speed speed p.u. p.u. with with a reduction reduction certain certain fatigue fatigue loads loads as as an an additional additional power objective. objective. outputthis This controller controller wind turbine adjusts adjusts pitch pitch angle angle based based on on a comparison Optimization power wind turbine comparison ωm. The reference pitch angle βref ωm. The reference pitch angle is determined by PI controller. The operating rotational Prevention excessive rotational βref is speed determined and mechanical by PI power controller. above The rated operating wind. rotational speed speed must must be be limited limited between between and and p.u. p.u. The The pitch pitch angle angle control control a PMSG PMSG wind wind turbine turbine can can be be summarized summarized as as follows: follows: ω Optimization Optimization m, limit (1.2p.u.) power power output output wind wind turbine β turbine Prevention Prevention excessive excessive rotational rotational speed speed and and mechanical max β max mechanical power power above above βrated rated ref wind. ω wind. m ω m, m, limit limit ω m (1.2p.u.) (1.2p.u.) β min β max max Figure 6. Pitch angle controller. β min max max β ref ref β min min β min min Figure 6. Pitch angle controller. Figure 6. Pitch angle controller.

9 Appl. Sci. 2017, 7, KE Optimization Method Operation Scheme For effective utilization stored KE, it is essential to understand influenced operating point. The operation scheme that utilizes KE control as a fast reserve is adopted in proposed optimal strategy. In normal operation, a PMSG wind turbine acts in MPPT mode, which is most economically viable mode. However, to actively participate in frequency control, PMSG wind turbine operates at de-loaded power extraction curve. In event a severe frequency drop, de-loaded control mode is changed to KE reserve discharging mode with a trigger signal. The output P out wind turbine increases as P KE, and KE reserve mode maintains output command using de-loaded control during support time. The purpose proposed KE optimization method is to first, improve PMSG performance in terms frequency regulation capability and second, maximize KE reserve while determining de-loaded power reference for active power control wind farm. When operating under frequency support control, KE reserve obtained from individual PMSG wind turbines can be calculated using Equation (18). Assuming that wind farm consists N wind turbines operating at different wind speeds, total KE reserve wind farm can be expressed as KE res,w f = N ) 0.5J i (ωdel,i 2 ω2 m,opt,i i=1 (19) Here, ω del,i and ω m,opt,i are de-loaded rotational and optimal rotational speeds in i th PMSG wind turbine, respectively. The KE reserve mode can discharge a much greater amount KE than a conventional plant same rating and inertia. The rotational speed deceleration resulting from KE discharge can be calculated according to following equation. Jω m dω m dt ) = P m (P re f + P KE The active power control can immediately control active power output a PMSG wind generator (P out = P re f + P KE ) according to command determined in system. Because mechanical input P m wind energy in a turbine blade wind generator is only determined from given wind speed and rotational speed, P m is not instantly changed from active power control. Controlling output PMSG wind turbine generator results in rotational speed changing with difference between mechanical power input and active power output. If mechanical input and actual output wind turbine balance, rotational speed is stabilized and wind turbine operates at corresponding operation point. Figure 7 shows modified q-axis rotational current control loop for proposed frequency support control. The wind turbine would require a detection scheme to analyze frequency drop. The trigger signal for switching to KE reserve is obtained from variance in frequency. Figure 8 shows triggering scheme and frequency detection employed in proposed frequency support control. In general, when frequency system decreases because an accident, it is very difficult to determine scale actual accident or severity frequency problem until frequency is restored. A misjudgment by frequency detector may cause or frequency problems due to frequency support control PMSG wind turbine. To prevent this problem, frequency detector determines trigger signal by prioritizing frequency change rate system. All wind turbines simultaneously recognize frequency deviation using an event detector. The rate change frequency (ROCOF) [28] can be obtained as (20) d f system dt = 2H system P loss f nominal (21)

10 Appl. Sci. 2017, 7, Appl. Appl. Sci. Sci. 2017, 2017, 7, 7, Here, fsystem is a system frequency and fnominal is nominal frequency (60 Hz). Equation (21) Here, fsystem is a system frequency and fnominal represents Here, f system amount is system required frequency active and fpower is nominal is that can nominal nominal be computed frequency frequency using (60 Hz). (60 Hz). Equation Equation ROCOF (21) and (21) represents represents indicates security amount amount index required active power that can be computed using ROCOF and required active grid disturbance. power that can When be computed frequency using ROCOF system and continuously indicates indicates security index grid disturbance. When frequency system continuously decreases securitybelow index a frequency grid disturbance. standard, When frequency frequency detector continuously system continuously provides decreases trigger decreases below a frequency standard, frequency detector continuously provides trigger below signal ato frequency regulate standard, system frequency. detector continuously provides trigger signal to regulate signal to regulate system frequency. system frequency. Figure Figure Modified Modified q-axis q-axis rotational rotational current current control control loop. loop. Figure 7. Modified q-axis rotational current control loop. Figure The triggering scheme and frequency detection. Figure 8. The triggering scheme and frequency detection Optimal KE Solution Optimal KE Solution The KE optimization problem in this in this study study is solved is solved using using PSO [29], PSO which [29], is which adopted is adopted for optimal for tuning optimal The tuning KE KE optimization reserve KE value reserve problem for each value in wind this for farm. each study Maximizing wind is solved farm. using Maximizing additional PSO [29], KEwhich additional can be is translated adopted KE can into for be optimizing optimal translated tuning for into rotational optimizing KE speed reserve for in rotational value de-loaded for speed each control. wind The farm. de-loaded purpose Maximizing control. proposed additional The KE purpose optimization KE can be method translated proposed is to KE into maximize optimization optimizing KE method for reserve rotational is defined to maximize speed in Equation in KE (19). de-loaded reserve Here, defined because control. in The Equation optimal purpose (19). rotational Here, speed, proposed because ω m,opt,i KE optimal, optimization is a constant rotational method value speed, determined is to ωm,opt,i, maximize is from a constant wind KE value speed, reserve determined defined objective in from Equation function wind (19). speed, Here, KE optimization because objective function optimal problem rotational is defined KE optimization speed, as ωm,opt,i, problem is a constant is defined value as determined from wind speed, objective function KE optimization problem is defined as [ N N N 2 Max Max 0.5J 0 i J (ω i ( ω ) 2 Max 0.5J i ω del,i) ] 2, i (22) i=1 1 del, i (22) i= 1 The objective function is is aimed at ensuring that generators maximize ir KE reserve from The objective while function is aimed at active ensuring power that control generators support maximize ir frequency KE reserve from power stored rotor, while performing active power control to support frequency power system. stored system. rotor, For enhancing while performing frequency active support power capability control wind to support farms, frequency main issue this power For enhancing frequency support capability wind farms, main issue this paper realizes paper KE system. realizes For KE optimization enhancing method frequency to determine support capability de-loading wind power farms, value reference main issue for each this PMSG paper optimization method to determine de-loading power value reference for each PMSG systems [15,16], realizes systems KE [15,16], optimization and maximizes method to sum determine KE reserve de-loading value through power value Equation reference (22). When for each PMSG and maximizes sum KE reserve value through Equation (22). When wind farm operates wind systems farm operates [15,16], according and maximizes system sum operator s KE reserve output value command through Equation Pref,wf, (22). de-loaded When power wind according to system operator s output command P ref,wf, de-loaded power reference P del,i and farm reference operates according to system operator s output command Pref,wf, de-loaded power de-loadedpdel,i rotational and de-loaded speed ωrotational del,i speed individual ωdel,i PMSG individual wind turbine PMSG determined wind turbine determined control reference Pdel,i and de-loaded rotational speed ωdel,i in control system must satisfy following constraints. individual PMSG wind turbine determined system must satisfy following constraints. in control system must satisfy following constraints.

11 Appl. Sci. 2017, 7, Total power output constraints wind farm: P re f,w f = Power output constraints PMSG wind turbines: Rotational speed constraints PMSG wind turbines: N P del,i (23) i=1 P min,i P del,i P max,i (24) ω m,opt,i ω del,i ω m,limit (25) When wind turbines operate at proposed de-loaded control, output PMSG wind turbine is kept constant according to de-loaded power reference. Thus, sum de-loaded references individual wind turbine generators, P del,i should be equal to output wind farms according to output command system operator P ref,wf, which sets output wind farm for stable system operation. The minimum possible power is defined in range that can maintain stable wind power operation for a given wind speed. When a wind turbine is operated in de-loaded control mode, rotational speed reference must be determined within range constraints. From Equation (25), minimum rotational speed is limited from optimal rotational speed in MPPT control mode, and maximum rotational speed all wind turbines is set to rotational speed reference pitch angle control PSO Algorithm In this work, PSO is applied to find global solution to KE optimization problem in Equations (22) (25). Each feasible solution problem is termed a particle, which is assigned a random velocity and flown through problem space. The ith particle swarm in d-dimensional search space can be represented by its position x i and velocity v i. Each particle keeps track its previous best position (pbest) and best performing particle swarm is termed global best (gbest). According to particle s previous best experience and best experience among particles in its neighborhood, speed and direction particle velocity are updated [30,31]. Mamatically, this algorithm is formulated as follows: v t+1 id = wv t id + c 1r 1 (pbest t id xt id ) + c 2r 2 (gbest t id xt id ) (26) x t+1 id = xid t + vt+1 id, i = 1, 2,..., n (27) The variable pbestid t represents personal best position attained by d-dimensional quantity individual i from among its t times occurrence, and pbestid t denotes d-dimensional quantity best position attained globally by swarm. New position and velocity vectors are determined for each particle by combining its previous velocity with pbestid t and gbestt id (Equations (26) and (27)). The initialization process for velocities and positions particles is conducted by using vectors random numbers. Here, each wind turbine is considered to be a particle while whole complex wind turbines is modeling swarm. In or words, deloaded power outputs wind turbines are regulated as particles a swarm to achieve different possible answers for problem. It means that number wind turbines is equal to number particles swarm. Inertial weight w is modulus that controls impact previous velocity on current one, thus practically, an appropriate selection inertial weight is requested to optimize for diverse wind speeds and distribute across all operating areas wind farm. The inertia weight may be increased if objective values particles close to accordance or tend to local optimal. Conversely,

12 Appl. Sci. 2017, 7, inertia weight could be decreased. In this work, inertial weight calculation according to number iterations can be written as w(t) = w max w max w min t max t (28) Moreover, in order to permit algorithm to avoid falling into local minima and search in whole optimization space, and improve algorithm accuracy and convergence speed, acceleration coefficients in searching process are controlled as follows: c 1 = c 1,max c 1,max c 1,min t max t (29) c 2 = c 2,min c 2,max + c 2,min t max t (30) During updating, particle position x i and velocity v i are limited in available range. When maximum number iterations or performance index is reached, output gbest is global optimal solution Procedure Optimization Strategy The proposed strategy focuses on frequency regulation capability PMSG wind turbine. As shown in Figure 9, procedure for proposed optimal strategy is described as follows: 1. Obtain maximum power using MPPT and pitch angle control based on converter control. 2. Optimize KE reserve to improve frequency support capability PMSG wind turbine. The KE optimization problem is solved using PSO algorithm, which maximizes sum KE reserves to determine de-loaded control power reference for each wind farm. If number iterations is maximized, optimal decision value is used to maximize KE reserve. 3. Operate at de-loaded power point to maintain a certain power margin and ensure an additional KE reserve. The regulation margin is available to support system frequency in a transient duration. 4. Check change in system frequency caused by loss generators and disturbance. In addition, monitor frequency deviation power system. If frequency deviation problem (low frequency) occurs in power system, n go to Step 5. Orwise, if measured frequency is higher than nominal frequency, system operator provides a new command to de-loaded control and reduces output conventional generators. When power system is stable, return to Step In event a frequency disturbance, switch control trigger signal from de-loaded control to KE discharge control. The frequency regulation control utilizes additional reserves due to KE stored in PMSG wind turbine and rotor. 6. If operation in power system is stabilized from compensation, end procedure.

13 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, De-loaded control mode 4. Case Study 4.1. Data and Parameter Setting Figure 9. Proposed optimal strategy scheme. The validity proposed proposed optimal optimal control control and operation and operation strategy strategy was demonstrated was demonstrated by applying by it to applying a simulated it to power a simulated system, power as illustrated system, as in illustrated Figure 10. This in Figure system10. isthis composed system is rmal composed units, hydro rmal units, and hydro windunits, farms. and The wind rmal farms. units The are rmal combined units are withcombined a gas turbine with G2 a gas (50turbine MW) and G2 steam (50 MW) turbine and G1 steam (100turbine MW). The G1 hydro (100 MW). units have The hydro two hydro units turbines have two G3 hydro and G4, turbines rated at G3 70 MW and and G4, 60 rated MW, at respectively. 70 MW and The 60 load MW, is respectively. based on modeling The load to represent is based a on linear modeling frequency to dependency represent a on linear active frequency power dependency characteristics. on Wind active farms power WF1 WF5 characteristics. (each 16Wind MW) farms were linked WF1 WF5 to (each small 16 power MW) system were linked and an to individual small power wind farm system consists and an eight individual PMSG windfarm turbines consists (each 2 eight MW). PMSG To solve wind KE turbines optimization (each 2 MW). problem, To solve maximum KE optimization iteration number problem, in PSO maximum is set to 100, iteration and number population in size PSO is is set set to to , r 1 and r 2 arepopulation uniform random size is set factors 50. with r1 and assigned r2 are values uniform between random 0 and factors 1 in each with assigned values between 0 and 1 in each step and for each particle component to add randomness to velocity update. The inertial weight w is set to reduce from 0.9 to 0.4 in different iterations.

14 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, step and for each particle component to add randomness to velocity update. The inertial weight w Appl. Sci. 2017, 7, is set to reduce from 0.9 to 0.4 in different iterations. Each acceleration coefficients, c 1 and c 2, Each acceleration coefficients, c1 and c2, is set to between 0.5 and 2.5 to achieve a mean value is Each set to between 0.5 and 2.5 to achieve a mean value unity for stochastic factor [30]. Based on unity for acceleration stochastic coefficients, factor [30]. c1 and Based c2, on is set to characteristics between 0.5 and 2.5 to wind achieve turbine, a mean Figure value 11 characteristics depicts unity for correlated stochastic wind wind turbine, factor speed [30]. Figure priles Based 11 for depicts on individual characteristics correlated wind farms wind in speed wind small priles turbine, power for Figure individual system. 11 wind depicts These farms results incorrelated are small simulated wind power using speed system. priles PSCAD/EMTDC Thesefor results individual aretool simulated wind by adopting farms using in parameters PSCAD/EMTDC small power described system. tool in by adopting These Appendix results parameters are A. simulated described using in PSCAD/EMTDC Appendix A. tool by adopting parameters described in Appendix A. Figure 10. Simulated small power system. Figure 10. Simulated small power system. Figure 10. Simulated small power system. Figure 11. Wind speed variations at each wind farm. Figure 11. Wind speed variations at each wind farm. Figure 11. Wind speed variations at each wind farm Simulation Results 4.2. Simulation Results 4.2. Simulation The performances Results PMSG wind farms in three different cases are compared to verify effectiveness The performances proposed optimal PMSG wind strategy. farms in three different cases are compared to verify effectiveness The performances proposed optimal PMSG wind strategy. farms in three different cases are compared to verify Case 1: No frequency support control; effectiveness proposed optimal strategy. Case 1: 2: No Frequency frequency support control control; (5% de-loaded operation) without using KE optimization Case 1: No frequency support control; method; Case 2: Frequency support control (5% de-loaded operation) without using KE optimization Case 2: Frequency support control (5% de-loaded operation) without using KE optimization method; Case 3: Frequency support control using KE optimization method (proposed strategy). method; Case 3: Frequency support control using KE optimization method (proposed strategy).

15 Appl. Sci. 2017, 7, Case 3: Frequency support control using KE optimization method (proposed strategy). In Case 1, none wind farms utilize reserve margin for supporting regulations with frequency control and active power. The PMSG wind turbines are operated only at maximum aerodynamic efficiency and pitch angle is n retained at optimal value. In Case 2, PMSG wind turbines participate in frequency regulations using a fixed 5% de-loaded operation based on frequency support control without using KE optimization method. In Case 2, active power outputs from all wind farms are subjected to same power reduction according to a fixed de-loaded reference. To enhance frequency contribution PMSG wind turbines, Case 3 implements an additional KE optimization method to maximize total KE in rotor. In this case, each individual wind farm operates in de-loaded control mode to optimize additional KE generated from wind turbine under consideration. In this work, an initial system is considered where all synchronous units are operated in accordance with parameters listed in Table 1. Here, active load power is 270 MW. The steam turbine G1 maximizes output from power generation as a base generator in a small power system. The overall capacity wind farms represents approximately 28% total installed generation capacity. It is assumed that at t = 100 s, system operator requests power output reference for all wind farms to be approximately 40 MW. The columns in Table 2 represent wind speed, maximum power, optimal rotational speed, and de-loaded power reference individual wind farms in three cases. The power outputs in Case 2 and Case 3 satisfy power reference requirement specified by system operator. Table 1. Power output individual generators. Unit Generator Type Generation (MW) Capacity (MW) G1 Steam turbine G2 Gas turbine G3 Hydro turbine G4 Hydro turbine Total Table 2. De-loaded power output references for small power system. Wind Farm S w (m/s) ω m,opt (p.u.) P max (MW) Case 1 (MW) Case 2 (MW) Case 3 (MW) Total Table 3 presents a comparison results additional KE reserve values in wind farms for three cases. The additional KE in Case 3 can supply a greater power output to power system than KE in Case 2. When PMSG wind turbines operate at a near-rated wind speed, rotational speed is also restricted by maximum value generated from pitch angle control (1.2 p.u.). As shown in Table 3, rotational speed WF1 in Case 2 is limited to maximum rotational speed 1.2 p.u. (ω del ) through pitch angle control. Besides, rotational speed WF1 and WF5 in Case 3 are restricted by pitch angle control. The de-loaded power reference for Case 3 increases KE reserve considerably. The application output control using KE optimization method can enhance capability frequency regulation. To evaluate frequency contribution proposed optimal strategy, loss G2 is determined at t = 250 s in such a way that it promotes a change in frequency power system. When frequency drops owing to loss G2, wind farms are simulated to support frequency regulation in test system. Figures 12 and 13 illustrate variations in active power

16 Appl. Sci. 2017, 7, Table 3. Comparison KE reserves in Case 2 and Case 3. Appl. Sci. 2017, 7, Case 2 Case 3 Wind Farm ωdel (p.u.) KE (MWs) ωdel (p.u.) KE (MWs) output and 1 rotational speed in Case 3, respectively Since1.200 rotational speed WF1 and WF5 is limited to 1.22 p.u. by pitch angle controller, it is noted that rotational speed are appropriately controlled within 3 available range ( p.u.). Each wind farm operates in MPPT control mode after exhaustion KE reserve and fers additional power output to participate in frequency control during5 support time Total Table 3. Comparison KE reserves in Case 2 and Case 3. To evaluate frequency contribution proposed optimal strategy, loss G2 is determined at t = 250 s in such a way Case that 2it promotes a change Case in 3 frequency power Wind Farm system. When frequency drops ω del owing (p.u.) to KE loss (MWs) G2, wind ω del (p.u.) farms are KE simulated (MWs) to support frequency regulation in test system. Figures 12 and 13 illustrate variations in active power output and 2 rotational speed in Case , respectively Since rotational speed WF1 and WF5 is limited to p.u. by pitch angle controller, it is noted that rotational speed are appropriately controlled 4 within available range 9.760( p.u.) Each wind farm operates in MPPT control mode after exhaustion KE reserve and fers additional power output to Total participate in frequency control during support time. (a) (b) (c) Figure 12. Cont.

17 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, (d) (d) (e) (e) Figure 12. Active power for individual wind farms in Case 3. (a) WF1; (b) WF2; (c) WF3; (d) WF4; Figure 12. Active power for individual wind farms in Case 3. (a) WF1; (b) WF2; (c) WF3; (d) WF4; (e) Figure WF Active power for individual wind farms in Case 3. (a) WF1; (b) WF2; (c) WF3; (d) WF4; (e) WF5. (e) WF5. (a) (a) (b) (b) Figure 13. Cont.

18 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, (c) (d) (e) Figure 13. Rotational speed for individual wind farms in Case 3. (a) WF1; (b) WF2; (c) WF3; (d) WF4; Figure 13. Rotational speed for individual wind farms in Case 3. (a) WF1; (b) WF2; (c) WF3; (d) WF4; (e) WF5. (e) WF5. Figure 14 presents a comparison system frequency responses during a disturbance in three Figure cases. 14 When presents frequency a comparison drops because system disturbance, frequency responses minimum duringfrequency a disturbance responses in three Case cases. 1 and When 2 are frequency Hz drops and because Hz, respectively. disturbance, Then, minimum grid minimum frequency frequency responses Case 13 and is are Hz Therefore, Hz and Case , Hz, i.e., respectively. proposed Then, strategy, grid can minimum provide frequency greatest support Case 3to is improve Hz. Therefore, minimum Casegrid 3, i.e., frequency proposed among strategy, all can cases. provide It may also greatest help support in enhancing to improve robustness minimum grid frequency power system amongby alldecreasing cases. It may frequency also helpfluctuations in enhancingunder robustness a given load. power As shown system in Figure by decreasing 14, frequency frequency recovery fluctuations proposed under a strategy given load. is much As shown faster in than Figure that 14, Case frequency 1 and 2, recovery and hence, Case proposed 3 can provide strategymore is much power faster to than enhance that Case frequency 1 and 2, response and hence, power Case 3 system. can provide Case more 3 can power also accomplish to enhancea greater frequency development response power frequency system. support Case 3capability can also accomplish PMSG wind a greater turbines development using PSO-based frequency optimization support process capability for PMSG KE reserve. wind turbines Consequently, using it should PSO-based be noted optimization that process proposed foroptimal KE reserve. control Consequently, and operation it strategy should be can noted considerably that proposed enhance optimal ability control initial and and operation low-frequency strategy responses can considerably in short-term enhancefrequency abilityregulation. initial and low-frequency responses in short-term frequency regulation.

19 Appl. Sci. 2017, 7, Appl. Sci. 2017, 7, Figure 14. Comparison system frequency response. 5. Conclusions This study investigated performance an optimal control and operation strategy for PMSG wind turbines to improve ir frequency support capability in a power system. The PMSG wind generation system was was constructed constructed including including PMSG PMSG modeling modeling and integrated and integrated MSC and MSC GSCand controls. GSC controls. The PMSGThe wind PMSG turbine wind control turbine is implemented control is implemented using relevant using direct relevant and quadrature direct and component quadrature component controls on controls rotoron voltage, rotor implemented voltage, implemented using converters. using converters. The active The power active control power in this control study in this focused study onfocused MPPT on control MPPT to control extract to extract optimal optimal power power and and de-loaded de-loaded control control to obtain to obtain power power margin. margin. The KE The reserve KE reserve control control was also was applied also applied to enhance to enhance frequency frequency support capability. support capability. To improveto improve contribution contribution a wind turbine a wind to support turbine stable to support frequency stable regulation frequency viaregulation temporary via release temporary more release KE, KEmore optimization KE, method KE optimization was applied method using was PSO applied algorithm. using It can PSO be algorithm. observed from It can be case observed study from that results case study validate that proposed results validate strategy for proposed PMSGstrategy given for modeled PMSG frequency given variations. modeled frequency In proposed variations. optimal In strategy, proposed optimal frequency strategy, regulation frequency capability regulation PMSG capability wind turbine depends PMSG wind on turbine additional depends KE reserve on additional generated from KE reserve optimization generated from process and optimization powerprocess margin. and De-loaded power control margin. prevented De-loaded control wind system prevented instability wind that system may instability have resulted that from may excessive have resulted KE discharge. from excessive The results KE discharge. demonstrate The that results proposed demonstrate strategy that can proposed restrict strategy frequency can drop restrict while decreasing frequency drop disturbances while decreasing across overall disturbances operational across scale overall PMSG wind operational turbines. scale Consequently, PMSG wind turbines. proposed Consequently, optimal strategy proposed can significantly optimal strategy enhance can significantly system contribution enhance to frequency system contribution regulation capability to frequency wind turbines. regulation Furr capability research wind work turbines. will concentrate Furr inresearch solving optimization work will concentrate problem about in solving control optimization strategyproblem to account about for PMSG control wind strategy turbines, to compared account for with PMSG benchmark wind turbines, optimization compared algorithms. with benchmark optimization algorithms. Acknowledgments: This research was supported by Basic Science Research Program through National Acknowledgments: This research was supported by Basic Science Research Program through National Research Foundation Korea Korea (NRF) (NRF) funded funded by by Ministry Ministry Education Education (2017R1D1A1B ). (2017R1D1A1B ). This research This was also research supported was byalso Korea supported Electric Power by Korea Corporation Electric through Power Korea Corporation Electrical Engineering through & Korea Science Electrical Research Engineering Institute (grant & number: Science R15XA03-55). Research Institute (grant number: R15XA03-55). Author Contributions: Mun-Kyeom Kim contributed this manuscript solely. Author Conflicts Contributions: Interest: The Mun-Kyeom author declares Kim contributed no conflict this interest. manuscript solely. Conflicts Interest: The author declares no conflict interest. Nomenclature Nomenclature β Blade pitch angle β re f Reference value pitch angle Blade pitch angle β β max, β min Maximum and minimum values pitch angle ρ ref Reference Air density value (=1.205 pitch kg/m angle 2 ) βmax, λ βmin Maximum Tip speed and ratio minimum values pitch angle ρ λ Air density (=1.205 kg/m 2 opt Optimal tip speed ratio ) λ Tip speed ratio λ Optimal tip speed ratio opt

20 Appl. Sci. 2017, 7, B Friction coefficient generator C p Power coefficient wind turbine C p,opt Optimal power coefficient wind turbine S w Wind speed J Moment inertia wind turbine (kg/m 2 ) V dc DC link voltage V dc,ref Reference value DC link voltage V di, V qi Inverter voltages in d q axis V ds, V qs Stator voltages in d q axis V dg, V qg Grid-side voltages in d q axis I ds, I qs Stator currents in d q axis I ds,ref, I qs,ref Reference value stator currents in d q axis I dg, I qg Grid-side currents in d q axis I dg,ref, I qg,ref Reference values grid-side currents in d q axis θ r Electrical angle between a-axis and q-axis (rad) u Any variable (voltage, current, and flux linkage) to be transformed from abc frame to dq frame R s, R g Resistances PMSG winding and grid sides T e Electromagnetic torque T m Mechanical torque Ψ f Magnetic flux C Capacitance DC link capacitor L d, L q Machine inductances in d q axis L s, L g Inductances PMSG winding and grid sides p n Number pole pairs ω del Reference value rotational speed de-loaded control ω e Electrical rotational speed ω g Rotational speed grid side ω m Mechanical rotational speed ω m,limit Maximum limited rotational speed ω m,opt Optimal rotational speed ω m,ref Reference value mechanical rotational speed ω out Rotational speed output ω t Rotational speed turbine R Rotator radius P m Mechanical power P t, P g Active powers wind turbine and grid side P max, P min Maximum and minimum possible powers P margin Power margin P out Active power output P ref Reference value active power P del De-loaded power reference P ref,ke Reference value KE power P ref,wf Reference value active power wind farm N Number PMSGs in wind farm P KE KE power increment P loss Lacking active power under grid fault Q g Reactive power grid side Q ref Reference value reactive power K del Rotational speed coefficient de-loaded control K opt Coefficient rotational speed for maximum power KE res Kinetic energy (KE) reserve KE res,wf Total KE reserve wind farm H system Equivalent inertia entire power system

21 Appl. Sci. 2017, 7, f system Frequency power system f nominal Nominal frequency (60 Hz) f ref Reference value frequency vid t Velocity along dth dimension ith particle in iteration t xid t Current position along dth dimension ith particle in iteration t pbest, gbest Best positions for individual and swarm particles r 1, r 2 Random numbers between 0 and 1 c 1, c 2 Acceleration coefficients c 1,max, c 1,min Maximum and minimum value c 1 c 2,max, c 2,min Maximum and minimum value c 2 w Inertial weight w max, w min Maximum and minimum value inertial weight t Current iteration t max Maximum number iterations Appendix A. Simulation Parameters Table A1. Wind turbine parameters. Parameters Notation Value Blade radius R 38 m Air density ρ kg/m 2 Optimal power coefficient C p,opt Rated wind speed S w,rated 12 m/s Table A2. PMSG generator parameters. Parameters Notation Value Rated generator power P m,rated 2 MW Rated RMS line-to-line voltage V m,rated ρ 0.69 kv Rated machine speed ω m,rated rad/s Pole pairs P 11 Generator and turbine inertia constant H s PM flux Ψ f 136 Wb Stator d-axis inductance L md H Stator q-axis inductance L mq H Stator leakage inductance L si H Stator resistance R s 0.08 Ω Table A3. Steam turbine synchronous generator parameters. Notation Value Notation Value H 5.4 s X l p.u. X d p.u. X q p.u. X d p.u. X q p.u. X d p.u. X q p.u. T d s T q s T d s T q s

22 Appl. Sci. 2017, 7, Table A4. Gas turbine synchronous generator parameters. Notation Value Notation Value H 1.86 s X l 0.15 p.u. X d 1.94 p.u. X q 1.92 p.u. X d p.u. X q p.u. X d p.u. X q p.u. T d s T q s T d s T q s Table A5. Hydro turbine synchronous generator parameters. Notation Value Notation Value H 2 s X l 0.12 p.u. X d p.u. X q 1.95 p.u. X d p.u. X q p.u. X d 0.16 p.u. X q p.u. T d s T q s T d s T q s Table A6. PI parameters. Controller k p k i MSC GSC DC-link Pitch angle De-loaded References 1. Rodrigo, T.P.; Sílvio, F.R.; Edwin, W.; Ricardo, S.; Pavol, B.; Jan, P. Operation and Power Flow Control Multi-Terminal DC Networks for Grid Integration Offshore Wind Farms Using Genetic Algorithms. Energies 2013, 6, Alexander, K.; Bruno, U.S.; Lueder, V.B. Curtailment in a Highly Renewable Power System and Its Effect on Capacity Factors. Energies 2016, 9, 510. [CrossRef] 3. Mohamed, A.; Ralph, K. Fault-Ride through Strategy for Permanent-Magnet Synchronous Generators in Variable-Speed Wind Turbines. Energies 2016, 9, [CrossRef] 4. Hong, C.M.; Chen, C.H.; Tu, C.S. Maximum power point tracking-based control algorithm for PMSG wind generation system without mechanical sensors. Energy Convers. Manag. 2013, 69, [CrossRef] 5. Xiaolian, Z.; Can, H.; Sipeng, H.; Fan, C.; Jingjing, Z. An Improved Adaptive-Torque-Gain MPPT Control for Direct-Driven PMSG Wind Turbines Considering Wind Farm Turbulences. Energies 2016, 9, 977. [CrossRef] 6. Doherty, R.; Mullane, A.; Nolan, G.; Burke, D.J.; Bryson, A.; Malley, M.O. An assessment impact wind generation on system frequency control. IEEE Trans. Power Syst. 2010, 25, [CrossRef] 7. Ekanayae, J.; Holdsworth, L.; Jenkins, N. Control DFIG wind turbines. Power Eng. 2003, 17, [CrossRef] 8. Wu, Z.; Gao, W.; Wang, X.; Kang, M.; Hwang, M.; Kang, Y.C. Improved inertial control for permanent magnet synchronous generator wind turbine generators. IET Renew Power Gener. 2016, 10, [CrossRef] 9. Li, Y.; Xu, Z.; Wong, K.P. Advanced control strategies PMSG-based wind turbines for system inertia support. IEEE Trans. Power Syst. 2017, 32, [CrossRef] 10. Lalor, G.; Mullane, A.; O Malley, M. Frequency control and wind turbine technologies. IEEE Trans. Power Syst. 2005, 20, [CrossRef] 11. Morren, J.; de Haan, S.W.H.; Kling, W.L.; Ferreira, J.A. Wind turbines emulating inertia and supporting primary frequency control. IEEE Trans. Power Syst. 2006, 21, [CrossRef]

23 Appl. Sci. 2017, 7, Holdsworth, L.; Ekanayake, J.B.; Jenkins, N. Power system frequency response from fixed speed and doubly fed induction generator based wind turbines. Wind Energy 2004, 7, [CrossRef] 13. De Almeida, R.G.; Lopes, J.A.P. Participation doubly fed induction wind generators in system frequency regulation. IEEE Trans. Power Syst. 2007, 22, [CrossRef] 14. Ramtharan, G.; Ekanayake, J.; Jenkins, N. Frequency support from doubly fed induction generator wind turbines. IET Renew. Power Gener. 2007, 1, 3 9. [CrossRef] 15. Wilches-Bernal, F.; Chow, J.H.; Sanchez-Gasca, J.J. A fundamental study applying wind turbines for power system frequency control. IEEE Trans. Power Syst. 2016, 31, [CrossRef] 16. Pradhan, C.; Bhende, C.N. Adaptive deloading stand-alone wind farm for primary frequency control. Energy Syst. 2015, 6, [CrossRef] 17. De Almeida, R.G.; Castronuovo, E.D.; Lopes, J.P. Optimum generation control in wind parks when carrying out system operator requests. IEEE Trans. Power Syst. 2016, 21, [CrossRef] 18. Vidyanandan, K.V.; Senroy, N. Primary frequency regulation by deloaded wind turbines using variable droop. IEEE Trans. Power Syst. 2013, 28, [CrossRef] 19. Xue, Y.; Tai, N. Review contribution to frequency control through variable speed wind turbine. Renew. Energy 2011, 36, Diaz-Gonzalez, F.; Hau, M.; Sumper, A.; Gomis-Bellmunt, O. Participation wind power plants in system frequency control: Review grid code requirements and control methods. Renew. Sustain. Energy Rev. 2014, 34, [CrossRef] 21. Keung, P.K.; Li, P.; Banakar, H.; Ooi, B.T. Kinetic energy wind-turbine generators for system frequency support. IEEE Trans. Power Syst. 2009, 24, [CrossRef] 22. Kayikci, M.; Milanovic, J.V. Dynamic contribution DFIG-based wind plants to system frequency disturbances. IEEE Trans. Power Syst. 2009, 24, [CrossRef] 23. Hafiz, F.; Abdennour, A. Optimal use kinetic energy for inertial support from variable speed wind turbines. Renew. Energy 2015, 80, [CrossRef] 24. Zertek, A.; Verbic, G.; Pantos, M. Optimised control approach for frequency-control contribution variable speed wind turbines. IET Renew. Power Gener. 2012, 6, [CrossRef] 25. Zertek, A.; Verbic, G.; Pantos, M. A novel strategy variable-speed wind turbines participation in primary frequency control. IEEE Trans. Sustain. Energy 2012, 3, [CrossRef] 26. Chinchilla, M.; Arnaltes, S.; Burgos, J.C. Control permanent-magnet generators applied to variable-speed wind-energy systems connected to grid. IEEE Trans. Energy Convers. 2006, 21, [CrossRef] 27. Qiao, W.; Qu, L.; Harley, R.G. Control IPM synchronous generator for maximum wind power generation considering magnetic saturation. IEEE Trans. Ind. Appl. 2009, 45, [CrossRef] 28. Kundur, P. Power System Stability and Control; McGraw-Hill: New York, NY, USA, Kumar, M.; Perumal, N.; Irraivan, E. Advanced Pareto Front Non-Dominated Sorting Multi-Objective Particle Swarm Optimization for Optimal Placement and Sizing Distributed Generation. Energies 2016, 9, 982. [CrossRef] 30. Su, H.Y.; Hsu, Y.L.; Chen, Y.C. PSO-based voltage control strategy for loadability enhancement in smart power grids. Appl. Sci. 2016, 6, 1. [CrossRef] 31. Zeddini, M.A.; Pusca, R.; Sakly, A.; Mimouni, M.F. PSO-based MPPT control wind-driven Self-Excited Induction Generator for pumping system. Renew. Energy 2016, 95, [CrossRef] 2017 by author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under terms and conditions Creative Commons Attribution (CC BY) license (