Electric Power Systems Research 144 (2017) 233–242

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Modeling power networks using dynamic phasors in the dq0

reference frame
Yoash Levron a,∗ , Juri Belikov b
a
The Andrew and Erna Viterbi Faculty of Electrical Engineering, Technion—Israel Institute of Technology, Haifa 3200003, Israel
b
Faculty of Mechanical Engineering, Technion—Israel Institute of Technology, Haifa 3200003, Israel

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

Article history: The dynamic behavior of large power systems has been traditionally studied by means of time-varying
Received 26 July 2016 phasors, under the assumption that the system is quasi-static. However, with increasing integration of fast
Received in revised form renewable and distributed sources into power grids, this assumption is becoming increasingly inaccurate.
20 November 2016
In this paper, we present a dynamic model of general transmission and distribution networks that uses
Accepted 28 November 2016
dynamic phasors in the dq0 reference frame. The model is formulated in the frequency domain, and is
based on the network frequency dependent admittance matrix. We also present a simplified version of
Keywords:
this model that is obtained by a first-order Taylor approximation of the dynamic equations. The proposed
Quasi static
Average signals
models extend the quasi-static model to higher frequencies, while employing dq0 signals that are static
Dynamic phasors at steady-state, and therefore combine the advantages of high bandwidth and a well-defined operating
Inter-area oscillations point. The models are verified numerically using the 9-, 30-, and 118-bus test-case networks. Simulations
Multi-machine show that frequency responses of all models coincide at low frequencies and diverge at high frequencies.
Stability In addition, responses of the dq0 model in the time domain and in the abc reference frame are very close
to those of the transient model.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Dynamic processes occurring in large electric power systems are modeled at varying levels of abstraction [1,2]. In simple static models,
voltages and currents are assumed to be sinusoidal, and are represented by phasors. The network is typically described in this case by
the power flow equations. On the other extreme, transient models describe the system using nonlinear differential equations, which are
solved numerically in the time domain [3–8]. A third type of model, based on time-varying phasors, is called the average-value, dynamic
or quasi-static model [2]. This model is sufficiently accurate as long as phasors are slowly changing in comparison to the system frequency
[9,10]. A key advantage of the quasi-static model is that it enables long numeric step times, and hence can accelerate the simulation process.
In addition, since the quasi-static model employs phasors instead of sinusoidal AC signals, the system operating point is well-defined, a
property which considerably simplifies stability studies. Due to these properties, quasi-static models have been used extensively in the
analysis of dynamic interactions that occur in time frames of seconds to minutes, and have historically enabled studies of machines stability,
inter-area oscillation, and slow dynamic phenomena [1–3,11].
In recent years, with the emergence of small distributed generators and fast power electronics based devices, the assumption of quasi-
static phasors is becoming increasingly inaccurate. Due to these emerging technologies, voltage and current signals can contain high
harmonic components, and can exhibit fast amplitude and phase variations [9]. These developments have led to a new class of models
based on dynamic phasors. The concept of dynamic phasors has emerged from averaging techniques employed in power electronics, and
has been extended to three-phase systems [12]. Dynamic phasors generalize the idea of quasi-static phasors, and represent voltage and
current signals by Fourier series expansions in which the harmonic components are evaluated over a moving time window [13]. This
approach offers the benefits of a phasor based analysis, high accuracy, and fast simulations [14,15].
Due to these properties, dynamic phasors have been used in the analysis of synchronous and induction machines [12,16], HVDC sys-
tems [14,17,18], FACTS devices [10], sub-synchronous resonance [19,20], asymmetric systems [12,21], and asymmetric faults [16,22]. For

∗ Corresponding author.
E-mail addresses: [email protected] (Y. Levron), [email protected] (J. Belikov).

http://dx.doi.org/10.1016/j.epsr.2016.11.024
0378-7796/© 2016 Elsevier B.V. All rights reserved.
234 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242

Fig. 1. A unit in the network, showing signals in the abc reference frame.

instance, a dynamic phasor model of a line commutated HVDC converter is presented in [14]. The model represents the low-frequency
dynamics of the converter, and has lower computational requirements than a conventional transient model. A dynamic phasor represen-
tation of an unbalanced radial distribution system is presented in [21], where various components including a photovoltaic source and an
induction machine are modeled, and dynamic interaction of these components is shown. A simulator based on dynamic phasors has been
developed in [15,23], which uses differential switched-algebraic state-reset equations to describe the system components. The simulation
tool combines advantages of transient stability and electro-magnetic simulation programs, and is demonstrated on the IEEE 39 bus test-
case network. Work [24] extends the dynamic phasor concept to multi-generator, multi-frequency systems. The theory presented in [24]
enables study of electric power systems without assuming a single frequency. The proposed approach is validated on a twin-generator
system, and can be extended to larger networks. In addition, dynamic phasors are utilized in state estimation [25–29], in systems with
varying frequencies [30], and also in microgrids [31].
These recent works are mainly focused on modeling specific generators and loads, and do not provide a complete model of large
transmission networks. To bridge this gap, this paper presents a dynamic model of general transmission and distribution networks using
dynamic phasors in the dq0 reference frame, and shows how this model can be implemented in practice using time-domain state equations.
The first model presented in this work is formulated in the frequency domain, and is based on the network frequency dependent admittance
matrix. Since a direct numeric implementation of this model is challenging in general, we present a simplified version of it called the first-
order dq0 model, which is obtained by a first-order Taylor approximation of the dynamic equations. The proposed frequency-domain model
extends the quasi-static model to higher frequencies, while employing dq0 signals that are static at steady-state, and therefore combines
the advantages of high bandwidth and a well-defined operating point. The standard quasi-static model follows as a special case from the
model at low frequencies.
The paper continues as follows. Section 2 recalls basic concepts of dynamic phasors and explains how to formulate quasi-static models in
terms of dq0 quantities. Section 3 presents the proposed dq0 model. The first-order dq0 model is presented in Section 4. Section 5 provides
simple illustrative examples, followed by various numerical results provided in Section 6. Concluding remarks are drawn in Section 7.

2. Dynamic phasors in the dq0 reference frame

For systems in steady-state, the dq0 transformation maps sinusoidal signals to constant quantities. This transformation is compatible
with standard electric machine models, and with emerging models of renewable and distributed generators [10,15,32,33].
Assume a general power network containing N three-phase units, with voltages and currents as presented in Fig. 1.
The dq0 transformation (as defined in [26]) is given by
⎡  2
 2 ⎤
⎡ ⎤ cos(ωs t) cos ωs t − cos ωs t + ⎡ ⎤
xd (t) ⎢ 3 ⎥ xa (t)
3
⎢   ⎥
⎢ q ⎥ 2⎢ 2 2 ⎥ ⎢ b ⎥
⎣ x (t) ⎦ = 3 ⎢ − sin(ωs t) − sin ωs t − − sin ωs t + ⎥ ⎣ x (t) ⎦ , (1)
⎢ 3 3 ⎥
x0 (t) ⎣ ⎦ xc (t)
1 1 1
2 2 2
where ωs is the nominal system frequency, and x is replaced with either a voltage (vn ) or a current (in ) signal. A prominent feature of
the dq0 transformation is that for balanced and static systems, the direct and quadrature components xd , xq are constants, and x0 = 0. In a
quasi-static system xa , xb , xc are nearly sinusoidal, with slowly varying amplitudes and phases. In this case, the system can be modeled in
terms of time-varying phasors, which are defined in terms of the dq0 components as
1 q 1 q
Vn (t) = √ vdn (t) + jvn (t) , In (t) = √ ind (t) + jin (t) . (2)
2 2
Under the quasi-static approximation these dq components are identical to the xy components, as described in [34]. The following analysis
defines the standard quasi-static model and power flow equations in terms of dq0 signals. Quasi-static networks are usually described by
the admittance matrix Ybus . Define the vectors:
T T q q T q q T
V d (t) = vd1 (t), . . ., vdN (t) , I d (t) = i1d (t), . . ., iN
d
(t) , V q (t) = v1 (t), . . ., vN (t) , I q (t) = i1 (t), . . ., iN (t) ,
T T
V 0 (t) = v01 (t), . . ., v0N (t) , I 0 (t) = i10 (t), . . ., iN
0
(t) . (3)

Then the complex voltage and current phasors relate by

I d (t) + jI q (t) = Y bus V d (t) + jV q (t) (4)

 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242 235

and the real and imaginary parts of this equations are

       
I d (t) = Re Y bus V d (t) − Im Y bus V q (t), I q (t) = Im Y bus V d (t) + Re Y bus V q (t). (5)

These equations constitute a quasi-static model. Equivalently, quasi-static networks are also described by the power flow equations [33]


N

N
Pn (t) = |Vn (t)| |ynk ||Vk (t)| cos ∠ynk + ık − ın , Qn (t) = −|Vn (t)| |ynk ||Vk (t)| sin ∠ynk + ık − ın . (6)
k=1 k=1

In which the constants ynk are elements of the network admittance matrix Ybus , and powers, amplitudes and phases are formulated in
terms of dq0 signals as
  1 d   1 q
Pn (t) = Re Vn (t)In (t)∗ = v (t)ind (t) + vqn (t)inq (t) , Qn (t) = Im Vn (t)In (t)∗ = v (t)ind (t) − vdn (t)inq (t) ,
2 n 2 n
  
Vn (t) = √1 vdn (t)
2
+ vn (t)
q 2
,
q
ın (t) = ∠Vn (t) = atan2 vn (t), vdn (t) . (7)
2

If the system is not quasi-static, the phasors above can be generalized by dynamic phasors. These can be defined by a Fourier series
expansion, as in [16]. However, to simplify notations, we have chosen in this paper to define dynamic phasors with respect to the Fourier
transform as
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
Vnd (ω)  vdn (t) Ind (ω)  ind (t)
∞ ∞
⎢ q ⎥ ⎢ q ⎥ −jωt ⎢ q ⎥ ⎢ q ⎥ −jωt
⎣ Vn (ω) ⎦ = ⎣ vn (t) ⎦ e dt, ⎣ In (ω) ⎦ = ⎣ in (t) ⎦ e dt. (8)
−∞ −∞
Vn0 (ω) v0n (t) In0 (ω) in0 (t)

3. Large network dynamics in the dq0 reference frame

The power flow equations and the equivalent dq0 model in (5), (6) are only valid under the quasi-static approximation, assuming slow
variations in amplitudes and phases [34]. The following theorem extends the quasi-static model to higher frequencies, and describes the
system dynamics for general dq0 signals.

Theorem 1. In symmetric power networks, a dynamic model based on dq0 dynamic phasors can be described as
⎡ ⎤ ⎡ ⎤ ⎡ V d (ω) ⎤
I d (ω) U(ω) jR(ω) 0
⎢ q ⎥ ⎣ −jR(ω) U(ω) 0 ⎦⎢ ⎥
⎣ I (ω) ⎦ = ⎣ V q (ω) ⎦ . (9)
0 0 Y bus (ω)
I 0 (ω) V 0 (ω)

The following definitions are used in (9):

1 1
U(ω):= Y bus (ω + ωs ) + Y bus (ω − ωs ) , R(ω):= Y bus (ω + ωs ) − Y bus (ω − ωs ) , (10)
2 2
and
T T q q T q q T
V d (ω) = V1d (ω), . . ., VNd (ω) , I d (ω) = I1d (ω), . . ., INd (ω) , V q (ω) = V1 (ω), . . ., VN (ω) , I q (ω) = I1 (ω), . . ., IN (ω) ,
T T
V 0 (ω) = V10 (ω), . . ., VN0 (ω) , I 0 (ω) = I10 (ω), . . ., IN0 (ω) . (11)

Proof. Following the definition of the dq0 transformation in (1), voltages and currents can be expressed as sums of three elements

V (t) = V d (t) cos(ωs t) − V q (t) sin(ωs t) + V 0 (t),

a
I a (t) = I d (t) cos(ωs t) − I q (t) sin(ωs t) + I 0 (t). (12)

Similar expressions apply to phases b, c with a proper phase-shift of ±2/3. Transformation to the frequency domain using the modulation
property yields
1 q q
V a (ω) = V d (ω − ωs ) + V d (ω + ωs ) + jV (ω − ωs ) − jV (ω + ωs ) + V 0 (ω),
2
1 d q q
I a (ω) = I (ω − ωs ) + I d (ω + ωs ) + jI (ω − ωs ) − jI (ω + ωs ) + I 0 (ω). (13)
2
According to definition
T T
I1a (ω) ··· INa (ω) = Y bus (ω) V1a (ω) ··· VNa (ω) . (14)

Substituting (14) in (13) yields

q q
I d (ω − ωs ) + I d (ω + ωs ) + jI (ω − ωs ) − jI (ω + ωs ) + 2I 0 (ω)
q q
= Y bus (ω) V d (ω − ωs ) + V d (ω + ωs ) + jV (ω − ωs ) − jV (ω + ωs ) + 2V 0 (ω) . (15)
236 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242

Using relation (10), equation (15) can be rewritten as

I d (ω − ωs ) + I d (ω + ωs ) + jI q (ω − ωs ) − jI q (ω + ωs ) + 2I 0 (ω) = 2Y bus (ω)V 0 (ω) + U(ω − ωs )V d (ω − ωs ) + jR(ω − ωs )V q (ω − ωs )

+ U(ω + ωs )V d (ω + ωs ) + jR(ω + ωs )V q (ω + ωs ) + j(−jR(ω − ωs )V d (ω − ωs ) + U(ω − ωs )V q (ω − ωs )) − j(−jR(ω + ωs )V d (ω + ωs )

+ U(ω + ωs )V q (ω + ωs )), (16)

which after comparison of coefficients results in (9). 䊐

Note that Theorem 1 holds for symmetric networks, in which the matrix Ybus (ω) equally applies to each of the three phases. However,
the generators and loads connected to the network are not necessarily linear or symmetric.
The following corollary relates the proposed dq0 model to the standard quasi-static model.
∗
Corollary 1. When Ybus (ω + ωs ) = Ybus and Y bus (ω − ωs ) = (Y bus ) , equation (9) reduces to the quasi-static expressions in (5).

This corollary states that if the general Ybus (ω + ωs ) can be approximated by a constant matrix when ω → 0, then the dynamic model is
quasi-static.

4. The first-order dq0 model

Since the full dq0 model in (9) is, in general, hard to implement numerically, this section presents a simplified version of it called the
first-order dq0 model, which is obtained by a first-order Taylor approximation of the dynamic equations. In this case it is assumed that the
dq0 signals are bandlimited, and have significant energy only at low frequencies. This enables representation of the admittance matrix by
a Taylor series around ω = 0 as


M
ωk ∂Y bus (ω + ωs )
Y bus (ω + ωs ) ≈ |ω=0 ,
k! ∂ωk
k=0
 ωk ∂Y bus (ω − ωs )
M 
M  ∗
jk
k ∂Y bus (ω + ωs )
Y bus (ω − ωs ) ≈ |ω=0 , = (jω) |ω=0 , (17)
k! ∂ωk k! ∂ωk
k=0 k=0
 ωk ∂Y bus (ω)
M

Y bus (ω) ≈ |ω=0 ,

k! ∂ωk
k=0

∗
where in the expression for Ybus (ω − ωs ) we have utilized Y bus (ω) = (Y bus (−ω)) and employed nested functions derivatives. The standard
quasi-static approximation is obtained by a zero-order Taylor series with M = 0 such that
∗
Y bus (ω + ωs ) ≈ Y bus (ωs ), Y bus (ω − ωs ) ≈ Y bus (ωs ) . (18)

If these expressions are substituted into the original model (9), then the model reduces to the quasi-static equations in (5). A more accurate
dynamic model is obtained by the first-order Taylor series approximation (M = 1) as

∗ ∂Y bus (ω + ωs )
Y bus (ω + ωs ) ≈ Y bus (ωs ) + jωF, Y bus (ω − ωs ) ≈ (Y bus (ωs )) + jωF ∗ , Y bus (ω) ≈ Y bus (0) + jωF0 , F = −j |ω=0 ,
∂ω
∂Y bus (ω)
F0 = −j |ω=0 . (19)
∂ω
Substitution of these expressions in (9) yields
⎡ ⎤ ⎡ ⎤
I d (ω) V d (ω)
⎢ q ⎥ ⎢ ⎥
⎣ I (ω) ⎦ = (M2 + M1 jω) ⎣ V q (ω) ⎦ (20)

I 0 (ω) V 0 (ω)

with
⎡ ⎤ ⎡     ⎤
Re Y bus (ωs ) −Im Y bus (ωs ) 0
Re{F} −Im{F} 0
   
M1 = ⎣ Im{F} Re{F} 0 ⎦ , M2 = ⎢
⎣ Im Y bus (ωs ) Re Y bus (ωs ) 0
⎥
⎦. (21)
0 0 F0
0 0 Y bus (0)

This model can be implemented in practice using a linear state-space model

dx
= Ax + Bu
dt (22)
y = Cx + Du,
 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242 237

Fig. 2. Example 1: a network with a single inductor.

Fig. 3. Example 2: long transmission line feeding a matched load.

T T
where u = (V d (t), V q (t), V 0 (t)) , y = (I d (t), I q (t), I 0 (t)) . The matrices A, B, C, D are chosen such that at low frequencies the frequency
response of the state-space model (22) is equal to (20). The general frequency response is given by y(ω) = (C(jωI − A)−1 B + D)u(ω), and is
approximated at ω → 0 by

y(ω) ≈ C −A−1 I + jωA−1 B + D u(ω)

(23)
−1 −2
= −CA B + D − jωCA B u(ω).

Comparison of (23) and (20) provides

−CA−2 B = M1 , −CA−1 B + D = M2 . (24)

From (24), C, D can be computed for any full-rank matrices A, B as

C = −M1 B−1 A2 , D = M2 − M1 B−1 AB. (25)

The matrices A, B can be arbitrarily chosen. It is typically beneficial to select A, B such that D = 0, to obtain zero gain at ω → ∞. This criterion
results in

D = M2 − M1 B−1 AB ⇒ A = BM −1
1
M2 B−1 . (26)

In general, if the matrix M1 is not full-rank, then A can be computed by solving a standard least-squares minimization problem

A = BPB−1 , P = arg minM1 X − M2 22 . (27)

X

The matrix B can be arbitrarily chosen, for instance B = I.

5. Dynamic dq0 models: simple examples

The simple network in Fig. 2 demonstrates the difference between the quasi-static model and the full dq0 model.
In this example, the quasi-static model (5) provides

V d (ω) = −ωs LI q (ω), V q (ω) = ωs LI d (ω). (28)

In comparison, the dq0 model can be computed by (9), and results in

V d (ω) = jωLI d (ω) − ωs LI q (ω), V q (ω) = ωs LI d (ω) + jωLI q (ω), V 0 (ω) = jωLI 0 (ω). (29)

Note that when ω → 0 the quasi-static and the dq0 models are equivalent.
Another example (Fig. 3) is the long transmission line. In this example the dq0 dynamic phasors are shown to relate through time delays.
Assume a long power line of length l. The line is lossless, with
inductance Lx and capacitance Cx per unit length. It feeds a matched load with
an impedance equal to the characteristic impedance Zc = Lx /Cx . The resulting load current is delayed in respect to the source voltage
[33] as
1 abc
i2abc (t) = v (t − d), (30)
Zc 1

where the time delay is given by d = l Lx Cx .
238 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242

Fig. 4. Comparison of frequency responses in the 9-bus network for dq quantities from input 1 to output 4.

Fig. 5. Comparison of frequency responses in the 118-bus network for dq quantities from input 76 to output 77.

The corresponding dq0 model can be computed by (9). Defining Ybus (ω) = (1/Zc )e−jωd and applying the inverse Fourier transform, the
result is
1 q q 1 q 1 0
i2d (t) = cos(ωs d)vd1 (t − d) + sin(ωs d)v1 (t − d) , i2 (t) = − sin(ωs d)vd1 (t − d) + cos(ωs d)v1 (t − d) , i20 (t) = v (t − d). (31)
Zc Zc Zc 1

The direct and quadrature current components are delayed in respect to the source voltage.

6. Numerical results

In this section the various models are compared in terms of frequency and transient responses. Three models are tested: the full dq0
model, the first-order model, and the quasi-static model. These are compared to the full transient model, which is constructed in state-
space using Matlab’s Simscape Power Systems software package. The 9-, 30-, and 118-bus networks from the MatPower database [35] are
used.
 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242 239

Fig. 6. Comparison of the first 100 largest Hankel singular values of the dq0 model and the first-order model in the 118-bus network.

Fig. 7. Comparison of step-responses in the 9-bus system. The inputs (bottom plot) are steps in phases a, b, c at bus 1. The output is the current of phase a at bus 4.

An array of Bode plots representing the dq0, first-order, and quasi-static models is graphically illustrated in Figs. 4 and 5 for 9- and
118-bus networks. The plots depict the magnitude (in dB) and phase (in degrees) for several input/output pairs. The step input is applied
to a generator located at bus 1 and the output is measured at bus 4 for the case of 9-bus network. Similarly, for 118-bus network the step
input is applied to a generator located at bus 76 and the output is measured at bus 77. The figures show that the frequency responses
coincide at low frequencies (ω → 0) and diverge at high frequencies. Specifically, the quasi-static model is represented by a constant gain,
and has a constant gain and phase at all frequencies. The first-order model response coincides with the dq0 model response over a wider
frequency range and deviates only at high frequencies. These figures illustrate the main result presented in Theorem 1 that dq0 model
extends both quasi-static and first-order models by taking high-frequency dynamic phenomena into account.
Fig. 6 compares the largest singular values of the dq0 model and the first-order model for the 118-bus network. It can be seen that
the most energetic (largest) values are similar, meaning that the majority of energy is preserved. This in turn verifies the quality of the
first-order model.
Figs. 7–11 compare the transient response of the various models for the 9-, 30- and 118-bus networks. All inputs and outputs are three
phase signals. The transient model is simulated directly in the abc reference frame. The dq0 model, quasi-static model, and the first-order
model are represented in the dq0 reference frame, so input and output signals are converted accordingly using (1). To this end, the dq0
model is represented in the state-space form using implementation available in [36]. The first-order model is implemented using equations
(22)–(27).
Figs. 7 and 8 provide simulation results for the case of 9-bus network. In Fig. 7 the step input signals (shifted in time for each of the phases)
are applied to the generator 1 at bus 1 and the outputs are measured at bus 4. In Fig. 8 a more complicated scenario is considered. Before
the transient, the inputs are balanced three phase sinusoidal signals ua1 = sin(250t) at bus 1. Then, the transient is triggered by shorting
phase a to ground, i.e., ua1 = 0. Note that uc,d
1
= sin(250t ± 2/3). The output is the current of phase a measured at bus 4. The simulation
results confirm the above frequency domain analysis presented in Fig. 4 and the fact that the quasi-static model becomes inaccurate at
high frequencies. The similar settings were used to obtain simulation results for the 30- and 118-bus networks (see Figs. 9–11). The figures
show that transient responses of the dq0 models are similar to those of the transient models. The quasi-static model is implemented by
constant gains, and is inaccurate, except for balanced sinusoidal inputs. The first-order model is moderately accurate.
240 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242

Fig. 8. Comparison of transients in the 9-bus system. Before the transient, the inputs are balanced three phase sinusoidal signals at bus 1. The transient is triggered by
shorting phase a to ground. The output is the current of phase a at bus 4.

Fig. 9. Comparison of step responses in the 30-bus network. The inputs are steps at t = 0 in phases a, b, c. (a) input at bus 1, output at bus 2, (b) input at bus 5, output at bus
7, (c) input at bus 8, output at bus 28, (d) input at bus 11, output at bus 9.

Table 1
Errors in transient responses.

Fig. 7 Fig. 8(a) Fig. 10 Fig. 11(a)

−8 −9 −8
dq0 0.9595 × 10 0.9572 × 10 0.6557 × 10 0.9340 × 10−7
MSE fo 0.0035 3.4425 × 10−5 4.6116 × 10−5 4.9407 × 10−4
qs 0.1929 0.0020 0.0039 0.0170

dq0 0.7922 × 10−8 0.9706 × 10−9 0.3570 × 10−8 0.8491 × 10−7

ISE fo 0.0035 3.4446 × 10−5 1.1498 × 10−5 1.2363 × 10−4
qs 0.1913 0.0020 9.6554 × 10−4 0.0042
 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242 241

Fig. 10. Comparison of step-responses in the 118-bus system. The inputs (bottom plot) are steps in phases a, b, c of bus 76. The output is the current of phase a at bus 77.

Fig. 11. Comparison of transients in the 118-bus system. Before the transient, the inputs are balanced three phase sinusoidal signals at bus 76. The transient is triggered by
shorting phase a to ground. The output is the current of phase a at bus 77.

Errors in the models transient responses are presented in Table 1, where two typical metrics, the mean squared error (MSE) and the
integral squared error (ISE), are used. The errors are calculated in comparison to the transient model, which serves as a reference. It can be
seen that the magnitudes of errors for the dq0 model are significantly lower than those of first-order and quasi-static models.

7. Conclusion

Dynamic behavior and stability of large-scale power systems have been traditionally studied by means of time-varying phasors, under
the assumption that the system is quasi-static. However, with increasing integration of fast renewable and distributed sources into power
grids, this assumption is becoming increasingly inaccurate. This paper uses dynamic phasors in the dq0 reference frame to describe the
dynamics of large transmission and distribution networks. The proposed models do not employ the assumption of quasi-static phasors,
and therefore remain accurate over a wide range of frequencies. The full dq0 model is based on the frequency dependent admittance
matrix, and is approximated using the Taylor series expansion to generate models of lower complexity: the classic quasi-static model
is obtained by a zero-order approximation, while the more accurate first-order model is obtained by a first-order approximation. The
developed models combine two properties of interest. On one hand, they provide the advantages of the dq0 reference frame. Specifically,
242 Y. Levron, J. Belikov / Electric Power Systems Research 144 (2017) 233–242

the models employ nonrotating dq0 signals that are static at steady-state, and thus enable a well-defined operating point. On the other
hand, the proposed models improve the accuracy of the quasi-static model at high frequencies, enabling representation of fast dynamic
phenomena in networks that include small distributed generators and fast power electronics based devices. In particular, these properties
can contribute to accurate and low complexity stability analysis in such networks. Simulations show that the frequency responses of all
models coincide at low frequencies and diverge at high frequencies. In addition, responses of the dq0 model in the time domain and in the
abc reference frame are very close to those of the transient model.

Acknowledgment

The work was partly supported by Grand Technion Energy Program (GTEP) and a Technion fellowship.

References

[1] Y. Liu, K. Sun, Y. Liu, A measurement-based power system model for dynamic response estimation and instability warning, Electr. Power Syst. Res. 124 (2015) 1–9.
[2] L. Miller, L. Cibulka, M. Brown, A.V. Meier, Electric distribution system models for renewable integration: Status and research gaps analysis, Tech. rep., California
Energy Commission, CA, USA, 2013.
[3] P.M. Anderson, A.A. Fouad, Power System Control and Stability, John Wiley & Sons, 2008.
[4] C. Dufour, J. Mahseredjian, J. Bélanger, A combined state-space nodal method for the simulation of power system transients, IEEE Trans. Power Del. 26 (2) (2011)
928–935.
[5] U.N. Gnanarathna, A.M. Gole, R.P. Jayasinghe, Efficient modeling of modular multilevel HVDC converters (MMC) on electromagnetic transient simulation programs,
IEEE Trans. Power Del. 26 (1) (2011) 316–324.
[6] R. Majumder, Some aspects of stability in microgrids, IEEE Trans. Power Syst. 28 (3) (2013) 3243–3252.
[7] T. Nishikawa, A.E. Motter, Comparative analysis of existing models for power-grid synchronization, New J. Phys. 17 (1) (2015).
[8] R. Zárate-Mi nano, T.V. Cutsem, F. Milano, J.A. Conejo, Securing transient stability using time-domain simulations within an optimal power flow, IEEE Trans. Power
Syst. 25 (1) (2010) 243–253.
[9] M. Ilić, J. Zaborszky, Quasistationary phasor concepts, in: Dynamics and Control of Large Electric Power Systems, Wiley, New York, 2000, pp. 9–60.
[10] P.C. Stefanov, A.M. Stanković, Modeling of UPFC operation under unbalanced conditions with dynamic phasors, IEEE Trans. Power Syst. 17 (2) (2002) 395–403.
[11] R. Yousefian, S. Kamalasadan, A Lyapunov function based optimal hybrid power system controller for improved transient stability, Electr. Power Syst. Res. 137 (2016)
6–15.
[12] A.M. Stanković, S.R. Sanders, T. Aydin, Dynamic phasors in modeling and analysis of unbalanced polyphase A machines, IEEE Trans. Energy Convers. 17 (1) (2002)
107–113.
[13] S. Almer, U. Jonsson, Dynamic phasor analysis of periodic systems, IEEE Trans. Autom. Control 54 (8) (2009) 2007–2012.
[14] M. Daryabak, S. Filizadeh, J. Jatskevich, A. Davoudi, M. Saeedifard, V.K. Sood, J.A. Martinez, D. Aliprantis, J. Cano, A. Mehrizi-Sani, Modeling of LCC-HVDC systems using
dynamic phasors, IEEE Trans. Power Del. 29 (4) (2014) 1989–1998.
[15] T. Demiray, G. Andersson, L. Busarello, Evaluation study for the simulation of power system transients using dynamic phasor models, in: Transmission and Distribution
Conf. and Expo, Bogotá, Colombia, 2008, pp. 1–6.
[16] A.M. Stanković, T. Aydin, Analysis of asymmetrical faults in power systems using dynamic phasors, IEEE Trans. Power Syst. 15 (3) (2000) 1062–1068.
[17] C. Liu, A. Bose, P. Tian, Modeling and analysis of HVDC converter by three-phase dynamic phasor, IEEE Trans. Power Del. 29 (1) (2014) 3–12.
[18] H. Zhu, Z. Cai, H. Liu, Q. Qi, Y. Ni, Hybrid-model transient stability simulation using dynamic phasors based HVDC system model, Electr. Power Syst. Res. 76 (6-7) (2006)
582–591.
[19] M.C. Chudasama, A.M. Kulkarni, Dynamic phasor analysis of SSR mitigation schemes based on passive phase imbalance, IEEE Trans. Power Syst. 26 (3) (2011)
1668–1676.
[20] P. Mattavelli, A.M. Stankovic, G.C. Verghese, SSR analysis with dynamic phasor model of thyristor-controlled series capacitor, IEEE Trans. Power Syst. 14 (1) (1999)
200–208.
[21] Z. Miao, L. Piyasinghe, J. Khazaei, L. Fan, Dynamic phasor-based modeling of unbalanced radial distribution systems, IEEE Trans. Power Syst. 30 (6) (2015) 3102–3109.
[22] S. Chandrasekar, R. Gokaraju, Dynamic phasor modeling of type 3 DFIG wind generators (including SSCI phenomenon) for short-circuit calculations, IEEE Trans. Power
Del. 30 (2) (2015) 887–897.
[23] T.H. Demiray, Simulation of power system dynamics using dynamic phasor models. Ph.D. thesis, TU Wien, 2008.
[24] T. Yang, S. Bozhko, G. Asher, Multi-generator system modelling based on dynamic phasor concept, in: European Conf. on Power Electronics and Applications, Lille,
France, 2013, pp. 1–10.
[25] T. Bi, H. Liu, Q. Feng, C. Qian, Y. Liu, Dynamic phasor model-based synchrophasor estimation algorithm for M-class PMU, IEEE Trans. Power Del. 30 (3) (2015)
1162–1171.
[26] D.-G. Lee, S.-H. Kang, S.-R. Nam, Modified dynamic phasor estimation algorithm for the transient signals of distributed generators, IEEE Trans. Smart Grids 4 (1) (2013)
419–424.
[27] R.K. Mai, L. Fu, Z.Y. Dong, K.P. Wong, Z.Q. Bo, H.B. Xu, Dynamic phasor and frequency estimators considering decaying DC components, IEEE Trans. Power Syst. 27 (2)
(2012) 671–681.
[28] R.K. Mai, L. Fu, Z.Y. Dong, B. Kirby, Z.Q. Bo, An adaptive dynamic phasor estimator considering DC offset for PMU applications, IEEE Trans. Power Del. 26 (3) (2011)
1744–1754.
[29] J. Ren, M. Kezunovic, An adaptive phasor estimator for power system waveforms containing transients, IEEE Trans. Power Del. 27 (2) (2012) 735–745.
[30] T. Yang, S. Bozhko, J.M. Le-Peuvedic, G. Asher, C.I. Hill, Dynamic phasor modeling of multi-generator variable frequency electrical power systems, IEEE Trans. Power
Syst. 31 (1) (2016) 563–571.
[31] X. Guo, Z. Lu, B. Wang, X. Sun, L. Wang, J.M. Guerrero, Dynamic phasors-based modeling and stability analysis of droop-controlled inverters for microgrid applications,
IEEE Trans. Smart Grids 5 (6) (2014) 2980–2987.
[32] A.E. Fitzgerald, C. Kingsley, S.D. Umans, Electric Machinery, 6th ed., McGraw-Hill, New York, 2003.
[33] J.J. Grainger, W.D. Stevenson, Power System Analysis, McGraw-Hill, New York, 1994.
[34] T.V. Cutsem, C. Vournas, Voltage Stability of Electric Power Systems, Kluwer Academic Publishers, Norwell, MA, 1998.
[35] R.D. Zimmerman, C.E. Murillo-Sánchez, R.J. Thomas, MATPOWER: steady-state operations planning and analysis tools for power systems research and education, IEEE
Trans. Power Syst. 26 (1) (2011) 12–19.
[36] Y. Levron, J. Belikov, Toolbox for modeling and analysis of power networks in the dq0 reference frame, MATLAB Central File Exchange, 2016 http://www.mathworks.
com/matlabcentral/fileexchange/58702 (retrieved 15.08.16).