betaFeedback

WorksEngineering

PLL-Less Synchronization of Grid-Following Converters Using a Voltage-Driven Frenet Frame

Abstract

In grid-following converters, the synchronization path can transmit voltage disturbances to the current loop under weak or unbalanced grid conditions. This paper develops a Frenet-frame controller that derives its axes from the integral curve of the measured voltage at the point of common coupling. The tangential current governs active power, while the normal and binormal currents govern the transverse components of non-active power. Feedforward terms derived from the output filter model compensate curvature and torsion coupling, yielding identical nominal current dynamics across the axes. The controller operates without a phase-locked loop or sequence extraction. A reduced transverse model accounts for the finite bandwidth of geometric rate reconstruction and predicts a reduction in low-frequency coupling from voltage disturbances to current. Controller hardware-in-the-loop tests using real-time simulation of a three-phase four-wire converter show the shortest settling time after an active power reference step and the smallest peak power deviations during balanced voltage disturbances among the tested controllers. Under single-phase voltage unbalance, the method reduces peak binormal current and maximum phase-current total harmonic distortion relative to both benchmarks. It also settles to the commanded power in a weak-grid connection test where the phase-locked loop benchmark exhibits sustained oscillations.

Introduction

Grid-following (GFL) voltage-source converters are widely used as interfaces for renewable generation, battery energy storage, and other inverter-based resources because of their mature implementation and clear control structure [1, 2, 3]. Most implementations rely on phase-locked loop (PLL) synchronization and dqdq-frame current regulation [4, 5, 6]. Under balanced sinusoidal conditions, this structure transforms three-phase ac variables into approximately constant quantities in the synchronous frame and thereby simplifies current and power regulation. However, PLL synchronization and the associated voltage processing become increasingly challenging under weak grids, voltage unbalance, waveform distortion, and rapid transients [7, 8, 9].

Synchronization and estimation structures based on the second-order generalized integrator (SOGI) have been developed for nonideal voltage conditions, including SOGI-based PLLs [10] and dual second-order generalized integrator frequency-locked loops (DSOGI-FLLs) [11, 12]. These methods improve synchronization-variable and voltage-component estimation under unbalance and distortion, but retain a dedicated voltage-processing stage between the measured grid voltage and current regulation.

A different research direction removes the explicit PLL from the GFL control structure. Voltage-modulated direct power control regulates instantaneous active and reactive powers directly in stationary coordinates [13, 14, 15]. Linear parameter-varying power-synchronized control (LPV-PSC) establishes synchronization through active-power dynamics [16, 17], while other PLL-less approaches employ oscillator-based synchronization, predictive control, or reactive-power synchronization [18, 19, 20]. However, removing the PLL alone does not eliminate the dynamic interaction between voltage processing, synchronization, and current regulation [21]. A further issue is therefore how the current-control coordinates and the associated power channels are constructed from the measured three-phase voltage.

Conventional GFL controllers generally employ stationary coordinates or prescribed planar rotating coordinates. Such representations are well suited to conventional active- and reactive-power regulation. Under general nonideal three-phase conditions, however, the integral curve generated by the three-phase voltage is not guaranteed to remain in a fixed plane. In a prescribed planar representation, additional components of the multidimensional instantaneous non-active-power vector generally remain functions of the same in-plane current variables used for conventional active- and reactive-power regulation. They are therefore not explicitly separated into distinct orthogonal current-control channels. This motivates a control representation generated directly from the instantaneous three-phase voltage trajectory that explicitly separates the multidimensional power components within the current-control coordinates.

Alternative coordinate and frequency representations have been developed for nonideal three-phase conditions. The Tan-Sun transformation [22] maps unbalanced ac quantities into dc quantities without second-harmonic components, whereas the generalized gg-dq0dq0 framework [23] provides time-invariant coordinates for prescribed phase- and magnitude-unbalanced waveforms. The complex-frequency framework [24] describes normalized voltage-magnitude and angular-frequency dynamics and has been further applied to converter synchronization and closed-loop estimation [25, 26]. These approaches provide useful representations of nonideal voltage dynamics, but they do not establish the voltage-trajectory-driven threedimensional coordinate representation and multidimensional power-channel structure considered in this paper.

The geometric-frequency framework provides a trajectory-based description of three-phase electrical quantities. In [27], frequency was interpreted through the local kinematics of three-phase trajectories in R3\mathbb{R}^{3}. The Frenet frame was subsequently shown to generalize the Park transform by defining tangent, normal, and binormal coordinates directly from the local trajectory without an externally specified synchronization angle [28], [29]. Geometric quantities have also been incorporated into control design [30], while the geometric interpretation of instantaneous power establishes a multidimensional representation of the non-active-power vector [31].

The three-phase voltage trajectory at the point of common coupling (PCC) defines a tangent-normal-binormal (TNB) frame. In this frame, the converter current components give distinct channels for active power and the two transverse non-active-power components. This paper develops the associated moving-frame model, decoupled current control, and analysis of converter–grid interaction.

The main contributions of this paper are summarized as follows.

  • A voltage-driven TNB current-control frame is constructed directly from the instantaneous three-phase PCC-voltage trajectory. The control axes require neither an explicit PLL synchronization angle nor positive-, negative-, or zero-sequence decomposition.

  • Combined with the geometric instantaneous-power representation, the TNB frame separates active-power transfer and the two transverse non-active-power components into distinct orthogonal current channels, providing a direct multidimensional power-regulation structure within the same moving frame.

  • The converter RLRL-filter dynamics are derived in the moving TNB frame, and curvature- and torsion-related coupling terms are explicitly compensated to obtain decoupled nominal current dynamics. A reduced transverse model is further developed to characterize converter–grid interaction around a balanced operating point.

  • Controller-hardware-in-the-loop (CHIL) tests compare the proposed method with PLL-based and PLL-less GFL benchmarks under matched plant, sampling, bandwidth, modulation, and current-limiting conditions. The tests cover reference changes, voltage disturbances, unbalance, and weak-grid operation.

The remainder of this paper is organized as follows. Section II develops the voltage-driven Frenet frame, TNB power representation, and moving-frame current controller. Section III develops the comparative dynamic and grid-interaction analysis. Section IV presents the experimental validation and benchmark comparisons. Section V concludes the paper.

Voltage-Driven Frenet-Frame Current Control

The controller uses a voltage-driven Frenet frame for GFL converters. Primes denote time derivatives, (⋅)T(\cdot)^{\mathsf{T}} denotes transpose, and ∥⋅∥\lVert\cdot\rVert is the Euclidean norm.

Voltage-Driven Frenet Coordinates

Let vp\mathbf{v}_{p} collect the instantaneous phase-to-neutral PCC voltages:

vp(t)≡vp,abc(t)=[vp,a(t)vp,b(t)vp,c(t)]T∈R3.(1)\mathbf{v}_{p}(t) \equiv\mathbf{v}_{p,abc}(t) = [v_{p,a}(t)\quad v_{p,b}(t)\quad v_{p,c}(t)]^{\mathsf{T}} \in\mathbb{R}^{3}. \tag*{(1)}

Following the geometric-frequency formulation, associate vp\mathbf{v}_{p} with the integral curve

xp(t)=xp(t0)+∫t0tvp(σ) dσ,xp′=vp.(2)\mathbf{x}_{p}(t) = \mathbf{x}_{p}(t_{0}) + \int_{t_{0}}^{t} \mathbf{v}_{p}(\sigma)\,\mathrm{d}\sigma,\qquad\mathbf{x}_{p}^{\prime} = \mathbf{v}_{p}. \tag*{(2)}

Thus, the tangent of xp\mathbf{x}_{p} is aligned with the instantaneous voltage, rather than with the locus traced directly by vp\mathbf{v}_{p}.

For twice-differentiable vp\mathbf{v}_{p} satisfying ∥vp∥>0\lVert\mathbf{v}_{p}\rVert> 0 and ∥vp×vp′∥≠0\lVert\mathbf{v}_{p}\times\mathbf{v}_{p}^{\prime}\rVert\ne0, the tangent, binormal, and normal unit vectors are defined as

T=vp∥vp∥,B=vp×vp′∥vp×vp′∥,N=B×T.(3)\mathbf{T} = \frac{\mathbf{v}_{p}}{\lVert\mathbf{v}_{p}\rVert},\qquad\mathbf{B} = \frac{\mathbf{v}_{p}\times\mathbf{v}_{p}^{\prime}}{\lVert\mathbf{v}_{p}\times\mathbf{v}_{p}^{\prime}\rVert},\qquad\mathbf{N} = \mathbf{B}\times\mathbf{T}. \tag*{(3)}

The triad {T,N,B}\{\mathbf{T},\mathbf{N},\mathbf{B}\} forms a right-handed orthonormal frame. Its time-domain curvature- and torsion-related rates are

ωκ=∥vp×vp′∥∥vp∥2,ωτ=vp⋅(vp′×vp′′)∥vp×vp′∥2∥vp∥.(4)\omega_{\kappa} = \frac{\lVert\mathbf{v}_{p}\times\mathbf{v}_{p}^{\prime}\rVert}{\lVert\mathbf{v}_{p}\rVert^{2}},\qquad\omega_{\tau} = \frac{\mathbf{v}_{p}\cdot(\mathbf{v}_{p}^{\prime}\times\mathbf{v}_{p}^{\prime\prime})}{\lVert\mathbf{v}_{p}\times\mathbf{v}_{p}^{\prime}\rVert^{2}}\lVert\mathbf{v}_{p}\rVert. \tag*{(4)}

Here, ωκ\omega_{\kappa} governs the local rotation within the osculating plane, whereas ωτ\omega_{\tau} describes the evolution of that plane. The Frenet transformation from abcabc to TNB coordinates is

F≜[TTNTBT],F−1=FT.(5)\mathbf{F} \triangleq\begin{bmatrix} \mathbf{T}^{\mathsf{T}} \\ \mathbf{N}^{\mathsf{T}} \\ \mathbf{B}^{\mathsf{T}} \end{bmatrix},\qquad\mathbf{F}^{-1} = \mathbf{F}^{\mathsf{T}}. \tag*{(5)}

For any vector y∈R3\mathbf{y}\in\mathbb{R}^{3},

yTNB=Fy,y=FTyTNB.(6)\mathbf{y}_{\mathrm{TNB}} = \mathbf{F}\mathbf{y},\qquad\mathbf{y} = \mathbf{F}^{\mathsf{T}}\mathbf{y}_{\mathrm{TNB}}. \tag*{(6)}

The Frenet–Serret kinematics give

F′=ΩFF,(7)\mathbf{F}^{\prime} = \boldsymbol{\Omega}_{\mathrm{F}}\mathbf{F}, \tag*{(7)}

where

ΩF=[0ωκ0−ωκ0ωτ0−ωτ0].(8)\boldsymbol{\Omega}_{\mathrm{F}} = \begin{bmatrix} 0 & \omega_{\kappa} & 0 \\ -\omega_{\kappa} & 0 & \omega_{\tau} \\ 0 & -\omega_{\tau} & 0 \end{bmatrix}. \tag*{(8)}

Since the tangential axis is aligned with the PCC voltage,

vp,TNB=Fvp,abc=[vp,T00],vp,T=∥vp,abc∥.(9)\mathbf{v}_{p,\mathrm{TNB}} = \mathbf{F}\mathbf{v}_{p,abc} = \begin{bmatrix} v_{p,T} \\ 0 \\ 0 \end{bmatrix},\qquad v_{p,T} = \lVert\mathbf{v}_{p,abc}\rVert. \tag*{(9)}

The converter-current vector is

iabc=[iaibic]T,(10)\mathbf{i}_{abc} = [i_{a}\quad i_{b}\quad i_{c}]^{\mathsf{T}}, \tag*{(10)}

and its TNB representation is

iTNB=Fiabc=[iTiNiB].(11)\mathbf{i}_{\mathrm{TNB}} = \mathbf{F}\mathbf{i}_{abc} = \begin{bmatrix} i_{T} \\ i_{N} \\ i_{B} \end{bmatrix}. \tag*{(11)}

where iTi_{T}, iNi_{N}, and iBi_{B} denote the tangential, normal, and binormal current components, respectively.

The PCC voltage thus determines both the frame and its geometric rates without an explicit synchronization angle.

Instantaneous Power Decomposition

Following the instantaneous-power formulation in [31], the scalar active power and vector non-active power are defined as

P=vp,abcTiabc,Q=vp,abc×iabc.(12)P = \mathbf{v}_{p,abc}^{\mathsf{T}}\mathbf{i}_{abc}, \qquad Q = \mathbf{v}_{p,abc} \times\mathbf{i}_{abc}. \tag*{(12)}

The vector QQ is referred to as the non-active-power vector to distinguish it from the scalar reactive power conventionally used under balanced operation.

Since F\mathbf{F} is orthonormal, the inner product is invariant under the coordinate transformation. Hence,

P=vp,TNBTiTNB=vp,TiT.(13)P = \mathbf{v}_{p,TNB}^{\mathsf{T}}\mathbf{i}_{TNB} = v_{p,T}i_T. \tag*{(13)}

Using

vp,abc=vp,TT,iabc=iTT+iNN+iBB,(14)\mathbf{v}_{p,abc} = v_{p,T}\mathbf{T}, \qquad\mathbf{i}_{abc} = i_T\mathbf{T} + i_N\mathbf{N} + i_B\mathbf{B}, \tag*{(14)}

the non-active-power vector becomes

Q=vp,T(iNB−iBN).(15)Q = v_{p,T}(i_N\mathbf{B} - i_B\mathbf{N}). \tag*{(15)}

Accordingly, two orthogonal components of QQ are

QB≜QTB=vp,TiN,Q_B \triangleq\mathbf{Q}^{\mathsf{T}}\mathbf{B} = v_{p,T}i_N,
QN≜−QTN=vp,TiB.(16)Q_N \triangleq-\mathbf{Q}^{\mathsf{T}}\mathbf{N} = v_{p,T}i_B. \tag*{(16)}

The signs in (16) follow from the orientation of the frame. The dot product retains the tangential current because the PCC voltage is aligned with T\mathbf{T}, while the cross product satisfies T×N=B\mathbf{T} \times\mathbf{N} = \mathbf{B} and T×B=−N\mathbf{T} \times\mathbf{B} = -\mathbf{N}. With the stated sign convention, PP, QBQ_B, and QNQ_N equal vp,Tv_{p,T} times iTi_T, iNi_N, and iBi_B, respectively, as illustrated in Figure 1. The control scheme in Figure 2 uses this correspondence to assign each power reference to its associated current channel. Proportional-integral (PI) regulators track the current references, and the resulting converter-voltage commands are implemented through pulsewidth modulation (PWM).

Figure 1. Geometric interpretation of instantaneous power in the voltage-driven TNB coordinates.

Figure 2. Block diagram of the proposed Frenet-frame current control scheme.

Moving-Frame Model and Decoupled Current Control

Each converter phase is connected to the PCC through an RLRL output filter with resistance RoR_o and inductance LoL_o. Let vg,abc\mathbf{v}_{g,abc} denote the converter-side phase voltages, expressed with the same reference as vp,abc\mathbf{v}_{p,abc}. For this symmetric per-phase filter model, the voltage balance in abcabc coordinates is

Loiabc′=vg,abc−vp,abc−Roiabc.(17)L_o\mathbf{i}'_{abc} = \mathbf{v}_{g,abc} - \mathbf{v}_{p,abc} - R_o\mathbf{i}_{abc}. \tag*{(17)}

From the product rule and (7), Fiabc′=iTNB′−ΩFiTNB\mathbf{F}\mathbf{i}'_{abc} = \mathbf{i}'_{TNB} - \mathbf{\Omega}_F\mathbf{i}_{TNB}. Premultiplying (17) by F\mathbf{F} then gives

LoiTNB′=vg,TNB−vp,TNB−RoiTNB+LoΩFiTNB.(18)L_o\mathbf{i}'_{TNB} = \mathbf{v}_{g,TNB} - \mathbf{v}_{p,TNB} - R_o\mathbf{i}_{TNB} + L_o\mathbf{\Omega}_F\mathbf{i}_{TNB}. \tag*{(18)}

Using vp,TNB=[vp,T 0 0]T\mathbf{v}_{p,TNB} = [v_{p,T}\ 0\ 0]^{\mathsf{T}}, the componentwise dynamics are

LoiT′=vg,T−vp,T−RoiT+ωκLoiN,LoiN′=vg,N−RoiN−ωκLoiT+ωτLoiB,LoiB′=vg,B−RoiB−ωτLoiN.(19)\begin{aligned} L_o i'_T &= v_{g,T} - v_{p,T} - R_o i_T + \omega_\kappa L_o i_N, \\ L_o i'_N &= v_{g,N} - R_o i_N - \omega_\kappa L_o i_T + \omega_\tau L_o i_B, \\ L_o i'_B &= v_{g,B} - R_o i_B - \omega_\tau L_o i_N. \tag*{(19)} \end{aligned}

Let vPI,Tv_{\mathrm{PI},T}, vPI,Nv_{\mathrm{PI},N}, and vPI,Bv_{\mathrm{PI},B} denote the outputs of the PI current regulators driven by iT∗−iTi_T^* - i_T, iN∗−iNi_N^* - i_N, and iB∗−iBi_B^* - i_B, respectively. The converter-voltage references are designed as

vg,T∗=vp,T+vPI,T−ωκLoiN,vg,N∗=vPI,N+ωκLoiT−ωτLoiB,vg,B∗=vPI,B+ωτLoiN.(20)\begin{aligned} v^*_{g,T} &= v_{p,T} + v_{\mathrm{PI},T} - \omega_\kappa L_o i_N, \\ v^*_{g,N} &= v_{\mathrm{PI},N} + \omega_\kappa L_o i_T - \omega_\tau L_o i_B, \\ v^*_{g,B} &= v_{\mathrm{PI},B} + \omega_\tau L_o i_N. \tag*{(20)} \end{aligned}

With exact feedforward and realizable voltage commands, (20) and (19) give identical nominal plants on the three axes:

Loik′=vPI,k−Roik,k∈{T,N,B}.(21)L_o i'_k = v_{\mathrm{PI},k} - R_o i_k, \qquad k \in\{\mathrm{T},\mathrm{N},\mathrm{B}\}. \tag*{(21)}

The frame derivative in (18) produces two coupling pairs. Curvature couples the tangential and normal channels, while torsion couples the normal and binormal channels. The feed-forward commands include the PCC voltage on the tangential axis and cancel LoΩFiTNBL_o\mathbf{\Omega}_F\mathbf{i}_{TNB} across the three axes. The resulting identical nominal plants allow the same current regulator to be used on each axis.

Current- and Power-Loop Parameter Design

Each decoupled current channel in (21) has the plant 1/(Los+Ro)1/(L_o s+R_o). The PI regulator Gi(s)=Kp,i+Ki,i/sG_i(s) = K_{p,i} + K_{i,i}/s cancels the plant pole and sets the closed-loop bandwidth to ωci\omega_{ci}:

Kp,i=Loωci,Ki,i=Roωci.(22)K_{p,i} = L_o\omega_{ci}, \qquad K_{i,i} = R_o\omega_{ci}. \tag*{(22)}

The nominal closed-loop response is Gci(s)=ωci/(s+ωci)G_{ci}(s) = \omega_{ci}/(s+\omega_{ci}). All three controllers use fci=ωci/(2π)=1 kHzf_{ci} = \omega_{ci}/(2\pi) = 1\ \mathrm{kHz} in the comparative tests.

For p=[P QB QN]T\mathbf{p} = [P\ Q_B\ Q_N]^{\mathsf{T}}, linearization at the operating point (VT0,iTNB,0)(V_{T0},\mathbf{i}_{TNB,0}) gives

Δp=VT0ΔiTNB+iTNB,0Δvp,T.(23)\Delta\mathbf{p} = V_{T0}\Delta\mathbf{i}_{TNB} + \mathbf{i}_{TNB,0}\Delta v_{p,T}. \tag*{(23)}

The gain design holds vp,Tv_{p,T} at VT0V_{T0}. The outer-loop crossover is chosen to satisfy ωo≤0.1ωci\omega_o \le0.1\omega_{ci}, with fo=ωo/(2π)=100f_o=\omega_o/(2\pi)=100 Hz in the tests. For Go(s)=Kp,o(1+ωz,o/s)G_o(s)=K_{p,o}(1+\omega_{z,o}/s) with ωz,o=5\omega_{z,o}=5 rad/s, unity open-loop gain at ωo\omega_o yields

Kp,o=1VT0∣Gci(jωo)∣1+(ωz,o/ωo)2(24)K_{p,o}=\frac{1}{V_{T0}|G_{ci}(j\omega_o)|\sqrt{1+(\omega_{z,o}/\omega_o)^2}} \tag*{(24)}
Ki,o=ωz,oKp,o.K_{i,o}=\omega_{z,o}K_{p,o}.

Uppercase KK denotes a continuous-domain gain, and lowercase kk in Table II denotes a normalized digital signal processor (DSP) coefficient.

ParameterSymbolValue
Common plant and operating conditions
Grid voltage and frequencyVLLV_{LL}, f0f_0380 V, 50 Hz
DC-link voltageVdcV_{dc}1300 V
Switching and sampling frequenciesfswf_{sw}, fsf_s15 kHz, 15 kHz
Output-filter parametersRoR_o, LoL_o0.02 Ω\Omega, 4 mH
Nominal grid parametersRgR_g, LgL_g0.0464 Ω\Omega, 6.960 mH
Active-power referenceP∗P^*22 kW
Matched current- and power-loop settings
Outer-loop DSP PI coefficientskp,Pk_{p,P}, ki,Pk_{i,P}0.1, 0.01
Inner-loop DSP PI coefficientskp,ik_{p,i}, ki,ik_{i,i}30, 3
Current bandwidth and power crossoverfcif_{ci}, fof_o1 kHz, 100 Hz
Common rate-deviation bound∣Δf∣max⁡|\Delta f|_{\max}5 Hz
D-PLL controller
Reactive-power referenceQ∗Q^*0 kVAr
D-PLL code-domain PI gainskp,PLLk_{p,\mathrm{PLL}}, ki,PLLk_{i,\mathrm{PLL}}0.24, 2.4
PLL natural frequency and sequence bandwidthfn,PLLf_{n,\mathrm{PLL}}, fseqf_{\mathrm{seq}}20 Hz, 80 Hz
D-PLL damping ratioζPLL\zeta_{\mathrm{PLL}}0.9
LPV-PSC controller
Reactive-power referenceQ∗Q^*0 kVAr
Power-control bandwidthfPSCf_{\mathrm{PSC}}100 Hz
Frenet controller
Non-active-power referencesQB∗Q_B^*, QN∗Q_N^*0, 0 kVAr
Measurement-filter settingsfaf_a, nan_a1.5 kHz, 2
Differentiator cutofffff_f1 kHz
Geometric-rate bandwidthfdf_d100 Hz

Table II. Plant and controller parameters.

The second-order measurement filter attenuates switching ripple before differentiation while retaining the voltage dynamics required for reconstruction. The differentiator cutoff fff_f is kept above the geometric-rate bandwidth fdf_d to limit attenuation and phase lag in the reconstruction band, and below one tenth of the sampling frequency to limit high-frequency gain. A higher fdf_d speeds up reconstruction but passes more derivative noise to the feedforward terms. The settings in Table II place fdf_d one decade below fff_f and fcif_{ci}.

Comparative Dynamic and Grid-Interaction Analysis

This section compares the decoupled phase-locked loop (D-PLL) and Frenet controllers under matched plant and current-loop conditions. The reduced analysis examines in-plane transverse coupling around a balanced operating point, synchronization response, PLL-bandwidth sensitivity, and weak-grid interaction.

D-PLL Benchmark Model

The positive- and negative-sequence decoupling stage of the D-PLL benchmark is represented by the equivalent sequence-extraction filter

Gseq(s)=ωseqs+ωseq,(25)G_{\mathrm{seq}}(s)=\frac{\omega_{\mathrm{seq}}}{s+\omega_{\mathrm{seq}}}, \tag*{(25)}

where ωseq=2πfseq\omega_{\mathrm{seq}}=2\pi f_{\mathrm{seq}}. The PLL regulator is

GDPLL(s)=KpDPLL+KiDPLLs.(26)G_{\mathrm{DPLL}}(s)=K_{p}^{\mathrm{DPLL}}+\frac{K_{i}^{\mathrm{DPLL}}}{s}. \tag*{(26)}

For a phase-detector gain V0V_0, the proportional and integral gains set the second-order design targets before sequence-filter dynamics are included (Gseq=1G_{\mathrm{seq}}=1):

KpDPLL=2ζPLLωn,PLLV0,KiDPLL=ωn,PLL2V0,(27)K_{p}^{\mathrm{DPLL}}=\frac{2\zeta_{\mathrm{PLL}}\omega_{n,\mathrm{PLL}}}{V_0},\qquad K_{i}^{\mathrm{DPLL}}=\frac{\omega_{n,\mathrm{PLL}}^2}{V_0}, \tag*{(27)}

where ωn,PLL=2πfn,PLL\omega_{n,\mathrm{PLL}}=2\pi f_{n,\mathrm{PLL}} and ζPLL\zeta_{\mathrm{PLL}} are the design natural frequency and damping ratio, respectively. In the PLL-bandwidth sensitivity analysis, ζPLL\zeta_{\mathrm{PLL}} and the sequence-extraction bandwidth are held constant, while the proportional and integral gains are recalculated from (27).

Around a balanced operating point, the transverse voltage perturbation and estimated-angle perturbation satisfy

ΔeDPLL=Gseq(ΔV⊥−V0Δθ^),\Delta e_{\mathrm{DPLL}}=G_{\mathrm{seq}}\left(\Delta V_{\perp}-V_0\Delta\hat{\theta}\right),
Δθ^=GDPLLsΔeDPLL,\Delta\hat{\theta}=\frac{G_{\mathrm{DPLL}}}{s}\Delta e_{\mathrm{DPLL}},
DDPLL(s)=s+V0GDPLLGseq.(28)D_{\mathrm{DPLL}}(s)=s+V_0G_{\mathrm{DPLL}}G_{\mathrm{seq}}. \tag*{(28)}

Eliminating the locking error and estimated-angle perturbation yields

ΔDPLL(s)=s2(s+ωseq)s+KpDPLL+V0ωseq(KpDPLLs+KiDPLL).(29)\Delta_{\mathrm{DPLL}}(s)=\frac{s^2(s+\omega_{\mathrm{seq}})}{s+K_{p}^{\mathrm{DPLL}}+V_0\omega_{\mathrm{seq}}\left(\frac{K_{p}^{\mathrm{DPLL}}}{s}+K_{i}^{\mathrm{DPLL}}\right)}. \tag*{(29)}

Hence, the sequence extractor and angle-locking loop introduce additional synchronization modes into the D-PLL current-control path.

Reduced Transverse Transfer Functions

The two controllers use the same continuous-domain inner current regulator

Gi(s)=Kp,i+Ki,is,Di(s)=Los+Ro+Gi(s).(30)G_i(s)=K_{p,i}+\frac{K_{i,i}}{s},\qquad D_i(s)=L_os+R_o+G_i(s). \tag*{(30)}

Let I0I_0 denote the operating current magnitude. For D-PLL, the reduced transverse PCC-voltage-to-current response is

YDPLL⊥(s)=−GDPLLGseq(V0+LoI0s)DiDDPLL.(31)Y_{\mathrm{DPLL}}^{\perp}(s)=-\frac{G_{\mathrm{DPLL}}G_{\mathrm{seq}}(V_0+L_oI_0s)}{D_iD_{\mathrm{DPLL}}}. \tag*{(31)}

Thus, the sequence-extraction and angle-locking dynamics enter directly into the transverse current channel.

For the Frenet controller, consider a balanced planar operating trajectory with constant voltage magnitude V0V_0. A small angular perturbation satisfies

ΔV⊥=V0Δθ,Δωκ=sΔθ.(32)\Delta V_{\perp}=V_0\Delta\theta,\qquad\Delta\omega_{\kappa}=s\Delta\theta. \tag*{(32)}

The corresponding ideal transverse voltage-to-geometric-rate relation is therefore

Kκ,id(s)≜Δωκ(s)ΔV⊥(s)=sV0.(33)K_{\kappa,\mathrm{id}}(s)\triangleq\frac{\Delta\omega_{\kappa}(s)}{\Delta V_{\perp}(s)}=\frac{s}{V_0}. \tag*{(33)}

Let Df(s)=ωfs/(s+ωf)D_f(s)=\omega_fs/(s+\omega_f) and Gr(s)=ωd/(s+ωd)G_r(s)=\omega_d/(s+\omega_d) denote the differentiator and geometric-rate filter, with ωf=2πff\omega_f=2\pi f_f and ωd=2πfd\omega_d=2\pi f_d. With residual delay TdT_d, the corresponding transverse reconstruction model is

K^κ(s)=Gr(s)Df(s)e−sTdV0,(34)\hat{K}_{\kappa}(s)=\frac{G_r(s)D_f(s)e^{-sT_d}}{V_0}, \tag*{(34)}

and the reconstruction residual becomes

Eκ(s)=Kκ,id(s)−K^κ(s)=s−Gr(s)Df(s)e−sTdV0.(35)E_{\kappa}(s)=K_{\kappa,\mathrm{id}}(s)-\hat{K}_{\kappa}(s)=\frac{s-G_r(s)D_f(s)e^{-sT_d}}{V_0}. \tag*{(35)}

The resulting estimator-inclusive transverse admittance is

YF⊥(s)=−LoI0Eκ(s)Di(s).(36)Y_F^{\perp}(s)=-\frac{L_oI_0E_{\kappa}(s)}{D_i(s)}. \tag*{(36)}

The residual Eκ(s)E_{\kappa}(s) measures the difference between the ideal and reconstructed geometric-rate responses to the same transverse voltage perturbation. In (36), LoI0Eκ(s)L_oI_0E_{\kappa}(s) gives the corresponding coupling-voltage response, and Di(s)D_i(s) relates that voltage to the transverse current.

Equations (31) and (36) thus preserve the dominant synchronization and reconstruction dynamics of the two control architectures within the same reduced-order framework. The responses in Fig. 3 are normalized by the balanced operating-point admittance I0/V0I_0/V_0, where V0V_0 and I0I_0 denote the Euclidean magnitudes of the balanced three-phase voltage and current vectors, respectively. Under balanced operation, V0=V_0 = VLL,rmsV_{LL,\mathrm{rms}} and I0=3Iph,rmsI_0=\sqrt{3}I_{\mathrm{ph,rms}}, where the subscript rms denotes root mean square (RMS). At 22 kW and zero non-active power, VLL,rms=380V_{LL,\mathrm{rms}}=380 V gives V0=380V_0=380 V and I0=57.89I_0=57.89 A.

The nominal D-PLL uses fn,PLL=20f_{n,\mathrm{PLL}}=20 Hz, ζPLL=0.9\zeta_{\mathrm{PLL}}=0.9, and an 80-Hz sequence-extraction bandwidth. The Frenet implementation uses fd=100f_d=100 Hz, ff=1f_f=1 kHz, and one sampling period of residual digital delay.

As shown in Figure 3, the D-PLL controller exhibits a pronounced low-frequency transverse-admittance peak associated with the sequence-extraction and angle-locking dynamics. Over 1–100 Hz, its maximum normalized magnitude is approximately −9.12 dB, compared with −21.93 dB for the Frenet controller, corresponding to an approximately 12.8-dB reduction in the dominant low-frequency transverse coupling.

Figure 3. Normalized transverse output-admittance characteristics. (a) Magnitude. (b) Phase.

The improvement is frequency dependent. The two magnitude characteristics intersect near 99 Hz, close to the selected geometric-rate bandwidth. Below this range, the Frenet reconstruction residual remains small, whereas at higher frequencies the differentiator, geometric-rate filter, digital delay, and current-loop dynamics increasingly determine the response. The proposed controller therefore primarily suppresses the low-frequency synchronization interaction rather than reducing the transverse admittance over the entire frequency range.

The D-PLL angle-tracking channel is

CDPLL(s)=V0GDPLL(s)Gseq(s)DDPLL(s).(37)C_{\mathrm{DPLL}}(s)=\frac{V_0G_{\mathrm{DPLL}}(s)G_{\mathrm{seq}}(s)}{D_{\mathrm{DPLL}}(s)}. \tag*{(37)}

For the Frenet controller, an equivalent angular-response channel can be constructed from the reconstructed geometric rate as

CF(s)=Gr(s)Df(s)e−sTd.(38)C_{\mathrm{F}}(s)=\frac{G_r(s)D_f(s)e^{-sT_d}}{}. \tag*{(38)}

This channel provides an equivalent measure of the response time associated with geometric-rate reconstruction and does not represent an internal angle state of the Frenet controller. Figure 4 compares the two responses to the same 10° PCC-voltage angle step.

Figure 4. Equivalent synchronization responses to a 10° PCC-voltage angle step.

fn,PLLf_{n,\mathrm{PLL}} (Hz)Max. normalized ∣YDPLL⊥∣|Y^\perp_{\mathrm{DPLL}}| (dB)2% settling time (ms)
10−9.8577.9
20−9.1234.6
30−7.8718.2
50−4.6729.3

Table 1. Sensitivity to D-PLL bandwidth.

The nominal D-PLL response is underdamped and reaches the 2% settling band in approximately 34.6 ms. The equivalent Frenet response settles in approximately 6.4 ms, consistent with the first-order settling-time estimate 4/ωd=6.374/\omega_d=6.37 ms at fd=100f_d=100 Hz. The latter response therefore reflects the implemented estimator dynamics, including the finite reconstruction bandwidth and digital delay.

D-PLL-Bandwidth Sensitivity

The influence of the D-PLL bandwidth is assessed by varying its natural frequency while maintaining ζPLL=0.9\zeta_{\mathrm{PLL}}=0.9, fseq=80f_{\mathrm{seq}}=80 Hz, and unchanged current-loop parameters. Table I summarizes the maximum normalized transverse-admittance magnitude over 1–100 Hz and the corresponding 2% angle-tracking settling time.

Increasing the PLL natural frequency from 10 to 30 Hz shortens the angle-tracking settling time but simultaneously increases the transverse admittance. Further increasing the natural frequency to 50 Hz no longer improves the settling time because of the stronger interaction with the fixed sequence-extraction dynamics. The nominal 20-Hz setting therefore provides a practical compromise among angle-tracking speed, sequence filtering, and transverse coupling.

The results further indicate that the synchronization response cannot be characterized by the current-loop bandwidth alone. For the D-PLL controller, it is jointly shaped by the PLL and sequence-extraction dynamics, whereas for the Frenet controller it is governed by the derivative and geometric-rate reconstruction paths. These method-specific dynamics are therefore preserved in the subsequent CHIL evaluation, with the remaining converter and implementation conditions kept common.

Weak-Grid Interaction

The converter–grid interaction is assessed through

det⁡[I+Zg(s)Yo(s)]=0,(39)\det\left[I+Z_g(s)Y_o(s)\right]=0, \tag*{(39)}

where Yo(s)Y_o(s) and Zg(s)Z_g(s) denote the converter output-admittance and grid-impedance matrices, respectively. For the transverse scalar reduction used in the subsequent sweep of the short circuit ratio (SCR), (39) becomes

1+Zg⊥(s)Y⊥(s)=0.(40)1 + Z_g^{\perp}(s)Y^{\perp}(s) = 0. \tag*{(40)}

The stronger low-frequency transverse coupling of the D-PLL controller shown in Figure 3 becomes increasingly influential as the grid impedance rises. The Frenet controller removes the PLL angle-locking path but does not eliminate converter–grid interaction. Below the geometric-rate bandwidth, the interaction is dominated by the reconstruction residual; at higher frequencies, the differentiator, geometric-rate filter, digital delay, and current-loop dynamics become increasingly important.

The controller formulations are independent of online knowledge of RgR_g and LgL_g. Grid strength is therefore varied directly through the external grid impedance in the subsequent SCR sweep.

The reduced models characterize the local small-signal converter–grid interaction around a balanced operating point and assume a stiff dc link and the adopted RLRL output filter. Switching-frequency effects, current limiting, modulation saturation, protection actions, and trajectory-validity transitions fall outside the scope of these models. Section IV-C compares the critical SCRs predicted from the closed-loop pole trajectories with those identified in the CHIL tests.

CHIL Validation

Experimental Platform and Comparison Conditions

The CHIL setup in Figure 5 pairs a real-time digital simulator (RTDS) with a TMS320C6657 DSP. The case study uses a three-phase four-wire converter. The RTDS models the converter, output filter, and Thevenin grid. It sends PCC voltages and converter currents to the DSP, which returns voltage commands at fs=15f_s = 15 kHz, with Ts=1/fsT_s = 1/f_s. The grid impedance is Zg=Rg+jω0LgZ_g = R_g + \mathrm{j}\omega_0L_g, where ω0=2πf0\omega_0 = 2\pi f_0. Table II lists the test parameters.

RTDS-based CHIL platform

Figure 5. RTDS-based CHIL platform.

The experimental comparison includes the D-PLL-based dqdq controller, LPV-PSC, and the proposed Frenet controller.

All implementations use the same converter, grid, operating point, sampling and switching frequencies, measurement conditioning, modulation constraint, current limit, and disturbance conditions.

LPV-PSC retains the published gain-scheduled outer power-synchronization law [16], with the power-control bandwidth set to the common 100-Hz target. The method-specific sequence-extraction, gain-scheduled power-synchronization, and geometric-rate reconstruction dynamics are retained in their respective implementations.

The current- and power-loop parameters follow Section II-D. The Frenet estimator settings are listed in Table II. The D-PLL gains are obtained from (27). The common 5-Hz rate-deviation bound is implemented as a protective saturation and does not determine the nominal bandwidth of any controller.

Baseline Dynamic Comparison

Case 1: Power-Reference Tracking

The first case evaluates active-power tracking at SCR = 3. The reference is increased from 20 to 25 kW at t=1.0t = 1.0 s and restored to 22 kW at t=1.2t = 1.2 s, while the corresponding non-active-power references remain zero. Active-power settling time is evaluated using a 2% band around the final steady-state mean.

The non-active-power metric is ∣Q∣|Q| for D-PLL and LPV-PSC and QB2+QN2\sqrt{Q_{B}^{2}+Q_{N}^{2}} for Frenet. The synchronization-related deviation is reported in per unit (p.u.) as (f−f0)/f0(f-f_{0})/f_{0} for the frequency estimated by D-PLL and LPV-PSC and as (ωκ−ω0)/ω0(\omega_{\kappa}-\omega_{0})/\omega_{0} for Frenet.

Figure 6 shows that the phase-current amplitudes increase after the upward power step and return near their initial levels when the reference is restored. For Frenet, (13) relates this change in power to the tangential current.

Figure 6. Three-phase PCC-voltage and converter-current responses in Case 1.

As shown in Figure 7, all three controllers track both reference changes. D-PLL rises more slowly and briefly undershoots at recovery. The signed non-active-power excursions reverse direction between transitions, with QBQ_{B} dominant for Frenet. For the upward step, Table III reports settling times of 20, 10, and 6 ms for D-PLL, LPV-PSC, and Frenet, respectively. Frenet gives the smallest non-active-power excursion, 0.84 kVAr, compared with 1.12 kVAr for LPV-PSC and 2.05 kVAr for D-PLL. The peak normalized rate deviation is 0.011 p.u. for both PLL-less controllers and 0.033 p.u. for D-PLL.

Figure 7. Active-power, non-active-power, and synchronization-related rate responses in Case 1.

Case 2: Balanced Grid Disturbance

A balanced 20° voltage-angle step is applied at t=1.0t=1.0 s. The source voltage magnitude is then reduced from 1.0 to 0.8 p.u. at t=1.3t=1.3 s and restored at t=1.6t=1.6 s. The voltage-driven trajectory remains planar, so this case primarily distinguishes the synchronization and reconstruction dynamics.

Figure 8 shows that the current amplitudes return near their pre-disturbance levels after the angle step. During the sag, the phase-current amplitudes increase for all three controllers as the PCC voltage falls. The active power returns near its pre-sag level while the current remains elevated, consistent with the higher tangential current required by (13) at lower voltage. The current amplitudes decrease as the voltage recovers.

Figure 8. Three-phase PCC-voltage and converter-current responses in Case 2.

Figure 9 shows a pronounced D-PLL active-power overshoot after the angle step and its largest non-active-power and rate excursions at voltage recovery. Frenet active power rises briefly at sag onset and dips at voltage recovery. The torsion-related rate stays close to zero, consistent with the planar trajectory. Over the full sequence, Table III reports the smallest peak active- and non-active-power deviations for Frenet, at 1.45 kW and 1.10 kVAr, respectively. Both PLL-less controllers reach 0.026 p.u. peak normalized rate deviation, compared with 0.050 p.u. for D-PLL.

Figure 9. Active-power, non-active-power, and synchronization-related rate responses in Case 2.

Case 3: Single-Phase Voltage Unbalance

A 0.5-p.u. magnitude reduction is applied to phase A of the Thevenin-source voltage from t=1.0t=1.0 to 1.41.4 s, reducing its magnitude from 1.0 to 0.5 p.u., while the other two phase magnitudes remain unchanged. The resulting single-frequency unbalanced voltage trajectory remains approximately planar but becomes noncircular, producing sustained oscillatory power and TNB-current components.

As shown in Figure 10, all three controllers maintain bounded phase currents during the disturbance. Figure 11 shows the largest non-active-power excursions for D-PLL. For Frenet, QBQ_B dominates the non-active-power response, with a plotted peak ∣QN∣\lvert Q_N\rvert of about 0.22 kVAr. The normalized torsion-related rate ωT/ω0\omega_T/\omega_0 stays close to zero during the sag and varies mainly at its onset and clearance.

Figure 10. Three-phase PCC-voltage and converter-current responses under the single-phase voltage unbalance in Case 3.

Active-power, non-active-power, and synchronization-related rate responses in Case 3

Figure 11. Active-power, non-active-power, and synchronization-related rate responses in Case 3.

Figure 12 resolves all phase currents in the same voltage-driven TNB frame. Frenet gives the smallest peak ∣iB∣\lvert i_B\rvert, with a plotted value of 0.75 A. Table III reports Case 3 peaks over 1.0–1.4 s and RMS values over 1.15–1.35 s from the plotted traces. Over these intervals, the peak and RMS non-active-power magnitudes are 2.9 and 1.9 kVAr for Frenet, 6.4 and 2.9 kVAr for D-PLL, and 4.3 and 2.5 kVAr for LPV-PSC.

Currents of the three controllers resolved in the common voltage-driven TNB frame in Case 3

Figure 12. Currents of the three controllers resolved in the common voltage-driven TNB frame in Case 3.

Current total harmonic distortion (THD) over 1.15–1.35 s is 100∑h=2HIh2/I1100\sqrt{\sum_{h=2}^{H} I_h^2/I_1}, where IhI_h is the RMS current of harmonic hh and HH is the highest resolved order. Panel (a) of Figure 13 displays the second through fifteenth harmonics.

Figure 13. Current-quality comparison in Case 3. (a) Phase-A harmonic spectrum from the second to fifteenth orders. (b) THD of each phase current using all resolved harmonics.

The fifth and seventh harmonics dominate the displayed spectrum. The maximum phase-current THD is 4.25% for D-PLL, 3.05% for LPV-PSC, and 1.75% for Frenet.

Weak-Grid Stability Boundary

Grid strength is characterized with

SCR=VLL,b2SbRg2+(ω0Lg)2,(41)\mathrm{SCR} = \frac{V_{LL,b}^{2}}{S_b\sqrt{R_g^{2}+(\omega_0 L_g)^{2}}}, \tag*{(41)}

where VLL,bV_{LL,b} and SbS_b denote the line-to-line voltage and apparent-power bases, respectively. The nominal X/RX/R ratio is retained while RgR_g and LgL_g are scaled proportionally to obtain SCR ∈{5,3,2,1.5,1.2}\in\{5, 3, 2, 1.5, 1.2\}.

At each grid strength, an independent CHIL grid-connection test is performed for each controller to determine whether it can establish the P∗=22P^* = 22 kW steady operating point. No controller is redesigned or retuned for an individual SCR test point. A test point is classified as stable when grid connection is established, the phase currents remain bounded, and the active power enters and remains within the 2% band around 22 kW without sustained post-connection oscillation.

Figure 14 gives critical SCR values of 1.793 and 1.386 for the D-PLL and Frenet controllers, respectively. The corresponding critical oscillation frequencies are approximately 32.6 and 587.7 Hz. The lower critical SCR of the Frenet controller indicates an extended weak-grid stability range. The limiting mode shifts from the low-frequency PLL-dominated interaction toward a higher-frequency range increasingly governed by the estimator, digital delay, and current-loop dynamics.

Figure 14. Closed-loop pole trajectories from the reduced weak-grid models under decreasing SCR. (a) D-PLL-based control. (b) Frenet control.

CaseTest conditionPerformance indexD-PLLLPV-PSCFrenet
Case 1Upward active-power reference stepActive-power settling time20 ms10 ms6 ms
Case 1Upward active-power reference stepPeak non-active-power deviation2.05 kVAr1.12 kVAr0.84 kVAr
Case 1Upward active-power reference stepPeak normalized synchronization-related rate deviation0.033 p.u.0.011 p.u.0.011 p.u.
Case 2Balanced phase-angle step and voltage sagPeak active-power deviation1.85 kW1.51 kW1.45 kW
Case 2Balanced phase-angle step and voltage sagPeak non-active-power deviation1.89 kVAr1.45 kVAr1.10 kVAr
Case 2Balanced phase-angle step and voltage sagPeak normalized synchronization-related rate deviation0.050 p.u.0.026 p.u.0.026 p.u.
Case 3Single-phase voltage-unbalance disturbancePeak common-frame ∣iB∣|i_B|16.1 A12.4 A0.75 A
Case 3Single-phase voltage-unbalance disturbancePeak non-active-power magnitude6.4 kVAr4.3 kVAr2.9 kVAr
Case 3Single-phase voltage-unbalance disturbanceRMS non-active-power magnitude2.9 kVAr2.5 kVAr1.9 kVAr
Case 3Single-phase voltage-unbalance disturbanceMaximum phase-current THD4.25%3.05%1.75%

Table 3. Representative CHIL performance comparison.

Table 4 gives CHIL critical SCRs of about 1.8 for D-PLL and 1.4 for Frenet, close to the analytical values of 1.793 and 1.386, respectively. LPV-PSC exhibits a comparable experimental stability boundary of approximately 1.4. Over the stable test points, Frenet also retains shorter grid-connection settling times as the grid weakens.

ControllerGrid-connection settling time at each SCR (ms)Critical SCR
5321.51.2
D-PLL dqdq152032UU1.8
LPV-PSC10131927U1.4
Frenet8101423U1.4

Table 4. Grid-connection stability and active-power settling time under different SCRs. U denotes instability; the critical SCR is determined from a finer sweep.

The agreement among the pole trajectories, the response at SCR = 1.7 in Figure 15, and the CHIL sweep indicates that the reduced transverse model captures the dominant local weak-grid interaction around the tested operating point. At SCR = 1.2, all three controllers are unstable; lower SCR values are therefore omitted from the principal sweep.

Figure 15. CHIL grid-connection active-power responses at SCR = 1.7 for the D-PLL-based and Frenet controllers.

Conclusion

This paper develops a PLL-less synchronization and current-control method for GFL converters operating under weak and unbalanced grid conditions. The method constructs a TNB frame directly from the instantaneous PCC voltage. Aligning the tangent with the voltage assigns active power to the tangential current and the two non-active-power components to the normal and binormal currents.

The moving-frame filter model identifies the cross-axis voltage terms introduced by curvature and torsion. Feedforward compensation yields identical nominal plants on all axes, allowing the same inner-loop tuning. A reduced transverse model relates the residual from finite-bandwidth geometric-rate reconstruction to the current response. For a 10° voltage-angle step, the modeled equivalent synchronization response settles in 6.4 ms with Frenet and 34.6 ms with D-PLL. Over 1–100 Hz, Frenet reduces the peak normalized transverse admittance by 12.8 dB relative to D-PLL.

In CHIL tests, Frenet settles in 6 ms after the active-power reference step, compared with 10 ms for LPV-PSC and 20 ms for D-PLL. Frenet gives the smallest non-active-power excursion in this test and the smallest peak deviations in active and non-active power during the balanced angle step, voltage sag, and recovery test. During single-phase voltage unbalance, Frenet yields the lowest peak binormal current, 0.75 A, and the lowest maximum phase-current THD, 1.75%. Its RMS non-active-power magnitude is 1.9 kVAr, compared with 2.5 kVAr for LPV-PSC and 2.9 kVAr for D-PLL.

The model predicts critical SCRs of 1.386 for Frenet and 1.793 for D-PLL. The CHIL sweep finds comparable critical SCRs for Frenet and LPV-PSC, with shorter grid-connection settling times for Frenet at their common stable test points. At SCR = 1.7, Frenet establishes the 22-kW operating point while the D-PLL power response continues to oscillate. The pole-derived stability boundaries apply locally around the balanced operating point used for linearization.

References

  1. [1]F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview of control and grid synchronization for distributed power generation systems,” IEEE Trans. on Industrial Electronics, vol. 53, no. 5, pp. 1398–1409, 2006.DOI
  2. [2]S. Golestan, J. M. Guerrero, and J. C. Vasquez, “Three-phase PLLs: A review of recent advances,” IEEE Trans. on Power Electronics, vol. 32, no. 3, pp. 1894–1907, 2017.DOI
  3. [3]M. Cespedes and J. Sun, “Impedance modeling and analysis of grid-connected voltage-source converters,” IEEE Trans. on Power Electronics, vol. 29, no. 3, pp. 1254–1261, 2014.DOI
  4. [4]N. Jaalam, N. A. Rahim, A. H. A. Bakar, C. K. Tan, and A. M. A. Haidar, “A comprehensive review of synchronization methods for grid-connected converters of renewable energy source,” Renewable and Sustainable Energy Reviews, vol. 59, pp. 1471–1481, 2016.DOI
  5. [5]Z. Ali, N. Christofides, L. Hadjidemetriou, E. Kyriakides, Y. Yang, and F. Blaabjerg, “Three-phase phase-locked loop synchronization algorithms for grid-connected renewable energy systems: A review,” Renewable and Sustainable Energy Reviews, vol. 90, pp. 434–452, 2018.
  6. [6]P. Rodriguez, J. Pou, J. Bergas, J. I. Candela, R. P. Burgos, and D. Boroyevich, “Decoupled double synchronous reference frame PLL for power converters control,” IEEE Trans. on Power Electronics, vol. 22, no. 2, pp. 584–592, 2007.
  7. [7]ENTSO-E ICS Investigation Expert Panel, “Final report on the grid incident in Spain and Portugal on 28 April 2025,” European Network of Transmission System Operators for Electricity (ENTSO-E), Tech. Rep., Mar. 2026.
  8. [8]A. Derwiškadze, G. Frigo, and M. Paolone, “Beyond phasors: Modeling of power system signals using the Hilbert transform,” IEEE Trans. on Power Systems, vol. 35, no. 4, pp. 2971–2980, 2020.
  9. [9]J. D. Lara, R. Henriquez-Auba, D. Ramasubramanian, S. Dhople, D. S. Callaway, and S. Sanders, “Revisiting power systems time-domain simulation methods and models,” IEEE Trans. on Power Systems, vol. 39, no. 2, pp. 2421–2437, 2024.
  10. [10]M. Ciobotaru, R. Teodorescu, and F. Blaabjerg, “A new single-phase PLL structure based on second order generalized integrator,” in Procs of the 37th IEEE Power Electronics Specialists Conf., 2006, pp. 1–6.DOI
  11. [11]P. Rodríguez, A. Luna, M. Ciobotaru, R. Teodorescu, and F. Blaabjerg, “Advanced grid synchronization system for power converters under unbalanced and distorted operating conditions,” in IECON 2006 - 32nd Annual Conf. on IEEE Industrial Electronics, 2006, pp. 5173–5178.DOI
  12. [12]P. Rodríguez, A. Luna, R. S. Muñoz-Aguilar, I. Etxeberria-Otadui, R. Teodorescu, and F. Blaabjerg, “A stationary reference frame grid synchronization system for three-phase grid-connected power converters under adverse grid conditions,” IEEE Trans. on Power Electronics, vol. 27, no. 1, pp. 99–112, 2012.DOI
  13. [13]Y. Gui, X. Wang, H. Wu, and F. Blaabjerg, “Voltage-modulated direct power control for a weak grid-connected voltage source inverter,” IEEE Trans. on Power Electronics, vol. 34, no. 11, pp. 11 383–11 395, 2019.DOI
  14. [14]Y. Gui, Q. Xu, F. Blaabjerg, and H. Gong, “Sliding mode control with grid voltage modulated DPC for voltage source inverters under distorted grid voltage,” CPSS Trans. on Power Electronics and Applications, vol. 4, no. 3, pp. 244–254, 2019.DOI
  15. [15]Z. Gong, C. Liu, L. Shang, Q. Lai, and Y. Terriche, “Power decoupling strategy for voltage modulated direct power control of voltage source inverters connected to weak grids,” IEEE Trans. on Sustainable Energy, vol. 14, no. 1, pp. 152–167, 2023.DOI
  16. [16]M. Z. Mansour, M. H. Rav anji, A. Karimi, and B. Bahrani, “Linear parameter-varying control of a power-synchronized grid-following inverter,” IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 10, no. 2, pp. 2547–2558, 2022.DOI
  17. [17]M. Z. Mansour, M. H. Rav anji, A. Karimi, and B. Bahrani, “Transient stability analysis of power-synchronized PLL-less grid-following inverters,” IEEE Trans. on Energy Conversion, vol. 39, no. 3, pp. 1851–1864, 2024.DOI
  18. [18]M. A. Awal and I. Husain, “Unified virtual oscillator control for grid-forming and grid-following converters,” IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 9, no. 4, pp. 4573–4586, 2021.arxiv.org/abs/2007.06042
  19. [19]X. Dong and H. Li, “A PLL-less voltage sensorless direct deadbeat control for a Sic grid-tied inverter with LVRT capability under wide-range grid impedance,” IEEE Trans. on Power Electronics, vol. 38, no. 8, pp. 9469–9481, 2023.DOI
  20. [20]Z. Tang, T. Hu, and G. Todeschini, “PLL-free approach: Reactive power-based synchronization in grid-following inverters,” IEEE Trans. on Energy Conversion, vol. 41, no. 3, pp. 1813–1823, Sep. 2026.DOI
  21. [21]S. M. Hosseinizadeh, S. Oumi, H. Karimi, M. Karimi-Ghartemani, and K. L. Lian, “Comparison of PLL-based and PLL-less vector current controllers,” IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 10, no. 1, pp. 436–445, 2022.
  22. [22]G. Tan, J. Cheng, and X. Sun, “Tan–Sun coordinate transformation system theory and applications for three-phase unbalanced power systems,” IEEE Trans. on Power Electronics, vol. 32, no. 9, pp. 7325–7380, 2017.DOI
  23. [23]A. Tayyebi, W. Zhang, X. Huang, W. Jiang, D. Bormann, and M. Larsson, “A generalized time-domain framework for modeling and analysis of the unbalanced three-phase systems,” Electric Power Systems Research, vol. 235, p. 110809, 2024.DOI
  24. [24]F. Milano, “Complex frequency,” IEEE Trans. on Power Systems, vol. 37, no. 2, pp. 1230–1240, 2022.
  25. [25]X. He, V. Häberle, and F. Dörfler, “Complex-frequency synchronization of converter-based power systems,” IEEE Trans. on Control of Network Systems, vol. 12, no. 1, pp. 787–799, 2025.DOI
  26. [26]X. Quan, J. Tang, T. Xu, Z. Wu, W. Gu, S. Pan, and A. Q. Huang, “Complex frequency locked loop for estimations of RoCoV and frequency,” IEEE Trans. on Power Systems, vol. 40, no. 6, pp. 5476–5479, 2025.DOI
  27. [27]F. Milano, “A geometrical interpretation of frequency,” IEEE Trans. on Power Systems, vol. 37, no. 1, pp. 816–819, 2022.arxiv.org/abs/2105.07762
  28. [28]F. Milano, “The Frenet frame as a generalization of the Park transform,” IEEE Trans. on Circuits and Systems I: Regular Papers, vol. 70, no. 2, pp. 966–976, 2023.DOI
  29. [29]F. Milano, G. Tzounas, I. Dassios, and T. Kerçi, “Applications of the Frenet frame to electric circuits,” IEEE Trans. on Circuits and Systems I: Regular Papers, vol. 69, no. 4, pp. 1668–1680, 2022.arxiv.org/abs/2112.03633
  30. [30]F. Sanniti, G. Tzounas, R. Benato, and F. Milano, “Curvature-based control for low-inertia systems,” IEEE Trans. on Power Systems, vol. 37, no. 5, pp. 4149–4152, 2022.DOI
  31. [31]F. Milano, G. Tzounas, and I. Dassios, “Instantaneous power theory revisited with classical mechanics,” IEEE Trans. on Circuits and Systems I: Regular Papers, vol. 71, no. 12, pp. 6080–6093, 2024.arxiv.org/abs/2401.15129

Paper details

vlnc:W_vf5wzzdv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…
Contents