This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

Pseudo-Dynamic Network Modeling for PMU

Based State Estimation of Hybrid AC/DC Grids
Wei Li, Student Member, IEEE, Luigi Vanfretti, Senior Member, IEEE, and Joe H. Chow, Fellow, IEEE

Abstract—This article presents a PMU-based state estimation become less valuable for on-line applications, especially those
algorithm that uses a pseudo-dynamic network modeling ap- with real-time requirements.
proach. The pseudo-dynamic network model combines difference Therefore, this work proposes a pseudo-dynamic modeling
equations with static network equations. Then it applies the
weighted least squares algorithm to solve an over-determined approach that can improve the SE accuracy during transients
least squares estimation problem. The proposed method can without significantly increasing the SE’s computational bur-
improve state estimation accuracy during both steady state den.
and transient conditions without increasing the computational At the same time, this work aims to address how the
burden. In addition, the proposed modeling approach is applied
power electronics-based devices (e.g., flexible AC transmission
to networks containing both a STATCOM and a VSC-HVDC to
demonstrate how to develop and apply a pseudo-dynamic state system (FACTS) and voltage source converter (VSC)-based
estimation model. Case studies aim to illustrate and verify the HVDC links) can be included in PMU-based SEs. As their
performance of the proposed method under steady state and installations continue to increase worldwide, these technolo-
transient conditions. gies will play an important role in improving the system
Index Terms—pseudo-dynamic, PMU, state estimation, STAT- control ability and flexibility with the increased integration
COM, VSC-HVDC. of renewable energy. Therefore, their real-time performance
during transients needs to be monitored in order to make
I. I NTRODUCTION full use of these devices in on-line operations. To this end,
this work develops suitable pseudo-dynamic models that can
P MU-based state estimation (SE) is gradually becoming
a practical and effective approach for power system
monitoring and control [1] [2] [3]. Compared to the telemetry
represent power electronics-based devices for both steady state
and transient conditions.
data used in conventional SCADA systems, PMUs provide
measurements with a global time stamp and in a much higher B. Related works
reporting rate, which allows operators to enhance operational
Work on tracking [6] and dynamic [7] SEs started as early as
security and to perform post-contingency analyses [4]. In
in the 1970s. In [6] the most recent estimated state set is used
addition to the high resolution measurement data, PMUs
as a one-step prediction, which is not always valid particularly
directly measure the system states that are to be estimated,
when the system is under transient operating conditions.
thus allowing to develop a state estimation network model
Although [7] firstly introduces Kalman filter from aerospace
with relatively low nonlinearity for the estimation algorithm
applications to power system SE, adopting oversimplified
to solve as compared with conventional SE approaches [4].
models to describe state’s time evolution results in a relatively
limited estimation capability. Since 1990s, efforts on including
A. Motivations a predictive database are taken to enhance the SE process via
As renewable energy integration brings intermittent fluctu- developing appropriate system models or/and implementing
ations into power systems, SE algorithms will face the need probability analyses. Interested readers are referred to [8] that
to capture system dynamics in a faster and more flexible way, provides a comprehensive survey on this topic. To the authors’
which can be difficult for the conventional SEs. On the other knowledge, this is the first article that proposes to combine a
hand, most of the so-called dynamic SEs, or forecasting-aided pseudo-dynamic model with a PMU-based static SE algorithm,
SEs, are computationally demanding, which introduces delays i.e., weighted least squares (WLS).
in the estimation calculation cycle [5]. When the calculation Previous work on FACTS modeling for SEs has been con-
cycle is longer than the PMU data rate, the estimated results ducted in the past decade. [10]- [12] present FACTS models
for the conventional static SEs using SCADA data. [13]- [16]
W. Li is with the Department of Electric Power and Energy Systems,
KTH Royal Institute of Technology, Teknikringen 33, SE-100 44 Stockholm, propose to use PMU data to conduct SEs for the networks
Sweden. E-mail: wei3@kth.se. containing FACTS devices. All the above work has focused on
J. Chow and L. Vanfretti are with the Department of Electrical, Computer, the static FACTS models for the static SEs, while this article
and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180
USA. E-mail: chowj@rpi.edu, vanfrl@rpi.edu. proposes a pseudo-dynamic model for the STATCOM, as an
W. Li is funded by the SweGRIDS project. example of FACTS devices.
J. H. Chow and L. Vanfretti are supported in part by the Engineering HVDC modeling for SEs dates back to 1980’s, when [17]
Research Center Program of the National Science Foundation and the Depart-
ment of Energy under NSF Award Number EEC-1041877 and the CURENT first includes a classic HVDC link model into an AC system
Industry Partnership Program. SE. [18] presents a relatively simplified AC/DC converter

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

model and extends it into a multi-terminal DC (MTDC) model A. WLS formulation and measurement model
for SEs. [19] focuses on modeling classic HVDC links with The SE can be formulated as a nonlinear WLS problem
only PMU measurements. With the rapid development and with the objective of minimizing the discrepancy between the
expansion of VSC-based devices, their SE models have also measured and the estimated values of each phasor quantity. In
been studied. [20] and [21] introduce a basic VSC model and this work, the WLS problem is formulated as
a generic VSC-based MTDC model for SEs, respectively. [22]
n+m
combines SCADA data with PMU data for hybrid SEs of VSC- X
HVDC links. [23] presents a more detailed VSC-HVDC link min wi e2i , (1)
x
i=1
model for a PMU-only SE. This article proposes a pseudo-
dynamic model for VSC-HVDC links in order to represent where  
the dynamic behavior of VSC-based devices. h(x)
e= . (2)
ε
C. Contributions Here e ∈ Rn+m is the error vector and ei is the ith row;
This article proposes a pseudo-dynamic modeling approach and wi is the ith diagonal element of the weight matrix. m
for PMU-based SEs, that offers the following contributions: and n denote the number of measured quantities and network
• High accuracy: It establishes difference equations from
equations, respectively.
the dynamic model of certain components from both The error vector e contains two parts: 1) network model
AC and DC grids, and then combines them with the equations h(x) ∈ Rn ; 2) errors between the measurements and
static network equations. This process is named: pseudo- their corresponding states ε ∈ Rm . Ideally, network equations
dynamic network modeling. would assume a perfect model, i.e., h(x) = 0, which is
• Compatible modeling and implementation: The pseudo-
not the case in practice. Hence, this work considers the case
dynamic network model maintains the basic structure of when the network equations may contain modeling errors, and
the static model, which greatly reduces the workload of thus, weights based on the confidence in the model’s accuracy
re-composing network models when updating existing are assigned to them. For the second part, as PMUs provide
algorithms. telemetry of the system states directly, the errors are for the
• Fast computation: the algorithm used to solve the static
quantities such as bus voltage magnitude |V| and angle θ, line
SE problem (WLS) is directly applied to the pseudo- current magnitude |I| and angle δ, and even other user-defined
dynamic SE, which ensures that the computational speed states, depending on what the PMUs are measuring.
will be minimally affected. The advantage of using Eq. (2) lies in the flexibility of grant-
• Explicit representation of control modes: For devices with
ing different weights to different network model equations
time-varying control references, the pseudo-dynamic SE since they inherently have different accuracies due to disparate
is capable of taking their control modes into account, reliabilities of the model’s parameters.
which would be challenging for static-only SEs. Due to space limitations, readers are referred to the non-
linear WLS solution in [1]– [3], [19]. This article focuses
In addition, two power electronics-based devices are used
on the enhancements to the network modeling, which will be
to illustrate the proposed approach—static synchronous com-
presented in the following subsection.
pensator (STATCOM), as an example of a FACTS device, and
VSC-HVDC link.
This article is organized as follows. Section II presents the B. Network model
PMU-based SE algorithm and the pseudo-dynamic network When using PMU data, the network model h(x) can be
modeling. The static and pseudo-dynamic network models for formulated as linear functions. This process requires careful
the STATCOM and the VSC-HVDC link are introduced in selection on states [4]. If the complex bus voltages and
Section III and Section IV, respectively. Simulation results complex line currents are selected as states, h(x) is linear
in Section V perform a comparison between the proposed and the resulting estimation problem can be solved without
pseudo-dynamic SE to the basic static SE. And Section iterations. However, in this article, bus voltage and line current
VI compares their computation performance. Finally, some phasors are defined in polar coordinates, that is, in the form of
conclusions are drawn in Section VII. magnitudes and angles, leading to a nonlinear network model.
This choice is relevant and important in order for the SE
II. PMU- BASED STATE ESTIMATION ALGORITHM to inherently detect and correct phasor angle errors [3] [19]
This section presents the PMU-based SE algorithm that [24] due to time asynchronization [25]. As the magnitude and
can be applied for both static SE and pseudo-dynamic SE. angle of a phasor quantity are largely independent variables
The only difference that exits is in the network models: the in a PMU, the angle error will be a single variable in the
static SE network model only contains static equations; while polar coordinates instead of two error quantities in rectangular
the pseudo-dynamic SE network model combines difference coordinates, which facilitates formulating angle errors in a
equations that describe system dynamical properties with static straightforward fashion.
network equations. The network model is combined with a The network model for the static SE describes the system
measurement model and the WLS is applied to solve the over- topology and properties under normal operating conditions,
determined estimation problem. which can be considered to be in the quasi-stationary regime.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

However, when the system is under transient conditions, the III. STATCOM NETWORK MODEL
static network model for certain components may no longer This section introduces the static and pseudo-dynamic mod-
hold. If the models for such components are not replaced by els of a STATCOM. The static model utilizes the static
the models that can represent their dynamic behaviors, the SE V-I characteristic when the STATCOM is under the linear
error would be significantly large. In other words, when the operation range; while the pseudo-dynamic model includes
system enters a transient condition due to a perturbation, if the STATCOM’s control process to better reflect its dynamic
the system cannot restore a steady state before the next PMU performance.
snapshot comes, the static network model will conflict with
the PMU measurement, leading to an inaccurate SE solution.
A. STATCOM static model
Therefore, a new type network model, called here the
pseudo-dynamic network model, is proposed. It leverages the The STATCOM aims to control the voltage at the connected
existing body of the network model and includes the difference bus [26]. It has a linear voltage-current (V-I) relation when it
equations that describe the system dynamical properties. The is under steady state operation conditions. A typical terminal
following section explains how to formulate the pseudo- voltage versus output current characteristic is shown in Fig. 1.
dynamic network model. Hence, for static analysis purposes, a STATCOM can be
modeled as an equivalent impedance, which is represented
Typically, it is assumed that the power system as a con-
by the slope Xs . This characteristic can be determined by
tinuous dynamical system can be described by employing
two points, (0, V ref ) and (|I0 |, |V0 |). The latter point is any
differential equations. However, for PMU-based SE purposes,
steady operating point in the V-I curve as long as the V ref
telemetry is streamed discretely over fixed and synchronous
is not changed by the operator. Several PMU snapshots of the
time intervals. In this sense, a power system can be treated as
STATCOM voltage and current phasors allow to determine this
a discrete dynamical system, where difference equations are
linear V-I characteristic, as illustrated using real PMU data in
used to update the state variables in discrete time steps of the
Sec. V-A.
same size as the PMU data.
This discretization is similar to numerically solving differen-
Xs
V
tial equations, i.e. numerical integration, using Euler’s method;
where the states are updated when knowing the starting point Vref
and the slope at it, and the error can be made small if the step
size is small enough and the interval of computation is finite.
However, Euler’s method is insufficiently robust, and thus the Capacitive Inductive
Euler’s full-step modification, which belongs to a second-order I st cap
I max ind
I max I st
Runge-Kutta method, is used next.
The difference equations used in this article are formulated Fig. 1: The voltage-current characteristic of STATCOMs
as
Ts
bk ≈ xk−1 + (ẋk−1 + x
x b˙ k ), (3) Using the V-I characteristic, its static network model can be
2 formulated as [16]:
where Ts is the sample time (step size). ẋ = g(x) can be h(x) : |V | − Xs |Ist | − V ref , (5)
either a linear or nonlinear function of x, which is essentially
the differential equation of the continuous dynamical system. where V is the bus voltage; V ref is the preset voltage
The value of ẋk−1 is calculated by substituting xk−1 into reference which the STATCOM aims to maintain; and Ist is
g(x), which is denoted by g(x)k−1 . the equivalent current phasor generated by the STATCOM.
Equation (3) implies that the present value is calculated by Therefore, the states of an AC system with STATCOMs can
adding the average increment during the time interval to the be denoted as
previous value. In order to comply Eq. (3) with the generalized x = [|V|, |I|, θ, δ, |Ist |] ,
T
(6)
form of the network model equation h(x), it is rewritten as
where |V|, |I|, θ and δ are AC system states. More details of
Ts
AC network model can be found in [19].
bk − xk−1 −
h(xk ) : x g(x)k−1 + g(x)k . (4)
2
The foregoing is the procedure of pseudo-dynamic network B. STATCOM pseudo-dynamic model
modeling and Eq. (4) is a pseudo-dynamic network model, This subsection introduces the pseudo-dynamic network
which can be used to describe all different components with model of a STATCOM. It is used as an example of how to
dynamical behavior. While the majority of the literature con- develop a pseudo-dynamic network model for a controller.
siders the use of Kalman Filters or other types of observers The STATCOM’s control process can be represented by
for a similar purpose, this work chooses a simpler approach, the block diagram shown in Fig. 2. The output is the current
which only requires few PMU measurements close to a certain magnitude |Ist |, which vary the current flow at the connected
component and without the need of formulating a complex bus in order to change the reactive power flow and control
dynamic model. Two examples are presented in the following the bus voltage. Ist is perpendicular to the bus voltage phasor,
sections. hence its angle can be computed from the voltage angle.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

V ref 
K
V
I dc
 I st
V T  s 1 I Zt Zp
Vf Vv
PCC Vdc
Fig. 2: STATCOM control block diagram If Phase Iv
a :1 reactor
Bus i AC
filter
This controller model is given by
K ref 1
|I˙st | =
(V − |V |) − |Ist |. (7)
T T Fig. 3: VSC substation structure
Using the pseudo-dynamic network model described by
Eq. (4), Eq. (7) can be written as:
Furthermore, in order to find the relation between con-
Ts b Ts K b Ts K ref
h(xk ) :(1 + )|Ist,k | + |Vk | − V verter’s voltage and current phasors with the DC voltage and
2T 2T T (8) current, the converter model needs to be included. For SE
Ts Ts K
− (1 − )|Ist,k−1 | + |Vk−1 |. purposes, an average value model (AVM) is sufficient for a
2T 2T VSC, which avoids distinguishing different switching levels
Equation (8) embodies the dynamic relation between |Vb | and modulation types, and instead it focuses on the funda-
and |Ibst |, and replaces the static relation in Eq. (5). This mental frequency voltage and current components. Therefore,
pseudo-dynamic network model of the STATCOM is validated the converter model is formulated as [23] :
using the real PMU data in Sec. V-A, and simulation studies 
in Sec. V-B.  Mv Vdc − |Vv |,

h(x) : 2 (11)
The states of the new model are the same as those of V
 (Kd2a )2 dc Idc − |Vv ||Iv | cos(θv − δv ),

the static model (Eq. (6)). The corresponding Jacobian matrix 2
elements must be updated using
where Mv is the modulation index, which is defined here as
∂h(xk ) Ts ∂h(xk ) Ts K the ratio of the root-mean-square (rms) value of the modulating
H(xk ) : =1+ , = . (9)
∂|Ibst,k | 2T ∂|Vbk | 2T wave, i.e., positive-sequence component of the AC voltage
at the converter, to the peak value of the carrier wave, i.e.,
IV. VSC-HVDC NETWORK MODEL the pole-to-pole DC voltage; Kd2a is the coefficient that
This section presents a VSC-HVDC substation’s static transfers DC quantities to the AC base when using per unit
model, its pseudo-dynamic model, and a VSC-HVDC link’s values; θv and δv are the converter voltage and current angles,
pseudo-dynamic model. The substation’s static model uses respectively.
the converter’s voltage to represent its voltage reference. The In summary, combining Eq. (10)-(11) gives a VSC substa-
reason why the voltage reference is not used as a state is tion model for static SE. The states of an AC system with
because it is internally generated by the control system, and VSC substations are given by
thus it cannot be considered in the static model. On the x = [|V|, |I|, θ, δ, |Vf |, θf , |If |, δf , |Vv |, θv , Vdc , Idc ] .
T
other hand, the pseudo-dynamic model includes the process (12)
of calculating this voltage reference, leading to a much more
accurate model. Then the substation’s pseudo-dynamic model
is extended to a point-to-point VSC-HVDC link model by B. VSC substation pseudo-dynamic model
including a DC link model. In Eq. (11), the modulation index Mv defines the relation
between Vdc and |Vv |. However, in reality |Vv | varies with the
A. VSC substation static model system dynamics; while Mv is a fixed value that depends on
the control mode. Therefore, in order to improve the accuracy
Figure 3 shows a VSC substation structure: V, I are the
of the model in Eq. (10)-(12) especially during transients,
voltage and current phasors at Bus i, which is also the point
|Vv | in Eq. (11) is replaced by |v ref |, which is the converter
of common connection (PCC); Vf is the voltage phasor at the
bridges’ voltage reference. This voltage reference is generated
AC filter and If is the current phasor flowing into it; Vv and
by the converter’s control system.
Iv are the voltage and current phasors at the AC side of the
The substation’s pseudo-dynamic model intends to repre-
converter; Zt and Zp are the impedances of the transformer
sent the control process of the converter. The most common
and phase reactor, respectively.
control strategy for VSCs is vector-current control, which
As indicated in Fig. 3, Bus i connects to the VSC substation
has a two-level control strategy, so-called the outer active-
through an AC line, which can be represented by the AC
reactive power and voltage loop (here abbreviated to outer
branch model to build up the relation between V, I, Vf , If
loop), and the inner current loop (here abbreviated to inner
and Vv , Iv . Hence, the static network model is formulated as
loop). As shown in Fig. 4, the outer loop transfers the VSC
[23] : ref
control references, i.e., P ref , Qref , Vdc ref
and Vac , into the

 Vv − Vf − Iv Zp , ref ref T
ref
h(x) : Vf − a1 V − aIZt , (10) converter’s current references, idq = (id iq ) . In this

Iv − If − aI. strategy, three-phase fundamental currents and voltages are

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

transformed into dq components in a synchronously rotating then

reference frame through Clark’s and Park’s transformations. Ts bref b ref

Hence, all quantities become DC signals [28]. y k = y k−1 + K i i − I v,k + i − I v,k−1 ,
2 k k−1
b
Depending on the converter’s operation mode, reference y ref ref ref
k−1 = −vk−1 + Vk−1 − Kp (ik−1 − Iv,k−1 ) − j(Xt + Xp )ik−1 .
iref
d can be determined by either active power P ref
or DC
ref Substituting ybk and yk−1 into Eq. (15), then
voltage Vdc . Similarly, reference irefq can be determined by
ref ref
either reactive power Q or AC voltage Vac at the PCC. h(xk ) :bvkref − vk−1 ref
− Vbk + Vk−1
For each VSC substation, only one iref d and only one i ref
q  
can be utilized. An ordinary integral controller can be used + Kp biref k − Ibv,k − iref
k−1 + Iv,k−1
for outer loop, which can be formulated using the pseudo- Ts bref b  (16)
dynamic equations (Eq. 4) as: + Ki ik − Iv,k + iref k−1 − Iv,k−1
2  
+ j(Xt + Xp ) biref − iref
!
ref ref .
ref ref T s P − Pbk P − Pbk−1 k k−1
h(xk ) :bid,k − id,k−1 − KP + ,
2 |Vf,k | |V f,k−1 |
As all the states at step (k − 1) are known, they can be
b
ref ref T s

ref

considered as a constant component and are denoted by C.
or id,k − id,k−1 − KV dc
b 2Vdc − Vdc,k − Vdc,k−1 ;
b
2 ! Hence, Eq. (16) can be simplified as
ref
biref − iref − KQ Ts Q −Q bk Qref − Q b k−1  
Ts bref b 
q,k q,k−1 + , h(xk ) :b ref
vk − Vk + Kp + Ki ik − Iv,k
2 |Vf,k−1 |
b
|Vbf,k | 2 (17)
ref ref Ts 
ref

+ j(X + X )i ref
+ C.
or iq,k − iq,k−1 − KV ac 2Vac − Vac,k − Vac,k−1 ; t p k
b b b
2
(13) The above equation can be split into d and q components:
where Pbk = |Vbk ||Ibk |cos(θbk − δbk ), Q b k = |Vbk ||Ibk |sin(θbk − δbk ).  
ref Ts bref
Next, the outputs of the outer loop become the inner h(xk ) :bvd,k − |Vbk |cos(θbk ) + Kp + Ki id,k −
2
loop’s input. The inner loop transfers the current references 
iref
dq = (id
ref ref T
iq ) into the voltage references of the converter |Ibv,k |cos(δbv,k ) − (Xt + Xp )biref q,k + Cd ,
ref ref ref T
bridges vdq = (vd vq ) , which then are transformed into   (18)
ref Ts bref
the three-phase voltage references. At last, a PWM algorithm v
b q,k − | V
b k |sin( θbk ) + K p + K i i q,k −
2
is utilized to generate switching commands for each sub- 
module. |Ibv,k |sin(δbv,k ) + (Xt + Xp )biref d,k + Cq .

Thus, the converter bridges’ voltage reference vref can be

Pref ref expressed by Eq. (13) and Eq. (18). So Eq. (11) becomes
I dref Vdref V a
Vdcref

 Kd2a Mv Vdc − (v ref )2 + (vqref )2
q
Vbref

d
Qref h(x) 2
Vdc (19)
I qref Vqref Vcref  (Kd2a )
 2
Idc − |Vv ||Iv | cos(θv − δv ).
Vacref 2
Outer loop control Inner loop control Transformation
Therefore, Eqs. (10), (19), (13), and (18) constitute the pseudo-
dynamic network model for VSC substations. As a result, four
Fig. 4: The vector-current control process for VSCs more states are added to this pseudo-dynamic model compared
to its static model:
Based on the converter’s structure shown in Fig. 3, the x =[|V|, |I|, θ, δ, |Vf |, θf , |If |, δf , |Vv |, θv , Vdc , Idc
converter bridges’ fundamental voltage equals to the voltage (20)
iref ref ref ref T
d , iq , vd , vq ] .
at PCC minus the voltage drop on the transformer and the
phase reactor. By neglecting the resistances of the transformer The Jacobian matrix needs to be updated w.r.t.
and the phase reactor, an ordinary PI controller can be used Eqs. (13), (18), and (19). Taking the first equation of
for the inner loop, which is formulated as: Eq. (13) as an example, its corresponding Jacobian matrix
1 element w.r.t. voltage magnitude is given by
v ref = V − (Kp + Ki )(iref − Iv ) − j(Xt + Xp )iref , (14)
s ∂h(xk ) Ts |Ibk |cos(θbk − δbk )
ref
H(xk ) : = KP .
where vb ref
= vbd + jb vqref and biref = biref
d + jbiref
q . ∂|Vbk | 2 |Vbf,k |
Equation (14) can be rewritten as:
C. Point-to-point VSC-HVDC link model
bv,k − Ki 1 biref − Ibv,k
   
vbkref =Vbk − Kp birefk − I
s k (15) Using the VSC substation model constructed above, other
ref VSC models can be developed. The point-to-point VSC-
− j(Xt + Xp )ik .
b
HVDC link is of primary interest here. To model the link,
In order to formulate Eq. (15) in the form of the pseudo- one VSC substation acts as the rectifier and the other one as
1
dynamic network model, assume that ybk = Ki (biref − Ibv,k ), the inverter, depending on which side controls the active power
s k

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

I cc as its output current phasor, was measured by a PMU with a

reporting rate of 30 samples/second.
Cdc Rdc Ldc Cdc
Figure 6 shows the STATCOMs V-I curve computed us-
I dc1 Vdc1 Vdc 2 I dc 2 ing the PMU data. Blue dots represent pre-fault operation
Cdc Rdc Ldc Cdc
points, which depict a linear V-I characteristic. When the
I cc fault occurred, several green dots scatter away from the V-
I characteristic. Gradually the system reached another stable
Fig. 5: DC circuit of a point-to-point VSC-HVDC link model operation point where the STATCOM current is about 0.25
p.u. Note that the STATCOM’s output current switched from
capacitive to inductive after the disturbance occurred, aiming
flow. At each substation a large DC capacitor is installed, and to decrease the bus voltage. When the system was approaching
a DC cable connects two substations [29]. The DC circuit of a the next steady state, the system operator changed the voltage
point-to-point VSC-HVDC link model is shown in Fig. 5 and reference of the STATCOM in several steps, which can be
the basic equations are: seen from the saw-tooth variation in red (the voltage reference

dVdc1 corresponds to the voltage when the STATCOM current output

 Cdc − (Idc1 − Icc ), equals to zero). The system reached a new operation point after

 dt
dVdc2

h(x) Cdc − (Idc2 + Icc ), (21) the reference changed, which is represented by the light blue
 dt dots. As the system turned to a new steady state condition, the
dI

 Ldc cc − (Vdc1 − Vdc2 − Rdc Icc ).

 system operator switched the STATCOM to the var reserve
dt control mode such that other slower voltage controls in the
The corresponding pseudo-dynamic equations are: system can take over the reactive power support.
Ts b Ts b
h(xk ) :Vbdc1,k − Idc1,k + Icc,k − Vdc1,k−1
2Cdc 2Cdc
Ts b Ts b
− Idc1,k−1 + Icc,k−1 ,
2Cdc 2Cdc
Ts b Ts b
Vbdc2,k − Idc2,k − Icc,k − Vdc2,k−1
2Cdc 2Cdc
Ts b Ts b
− Idc2,k−1 − Icc,k−1 ,
2Cdc 2Cdc
Ts Rdc b Ts b Ts b
(1 + )Icc,k − Vdc1,k + Vdc2,k
2Ldc 2Ldc 2Ldc
Ts Rdc Ts Ts
− (1 − )Icc,k−1 − Vdc1,k−1 + Vdc2,k−1 .
2Ldc 2Ldc 2Ldc
(22)
Fig. 6: The STATCOM PMU measurements and linear V-I
characteristic
V. C ASE STUDY
This section focuses on studying the performance of the Figure 6 shows the variation of operating conditions during
PMU-based SEs accuracy when using the static network model the whole event and clearly reveals the linear V-I droop
and the proposed pseudo-dynamic network model. Subsection relation of a STATCOM in steady-state operation. Using the
V-A uses real PMU data to compare and validate these two pre-fault PMU data (0 to 69.33s), this linear V-I characteristic
models in the case of a STATCOM. In Subsection V-B, two was determined by applying the linear regression function
test systems with a STATCOM installed were built up and polyfit in MATLAB, which is shown by the black line in
simulated in PSAT to generate the synthetic measurements for Fig. 6. The two coefficients of the linear predictor, slope and
the accuracy comparison. At last, the VSC-Based HVDC Link the intercept, represent the equivalent impedance and voltage
model provided by MATLAB/Simulink was used to generate reference of the STATCOM, respectively. For the pre-fault
synthetic measurements for two test scenarios used in the steady state, Xs = 0.0285 and V ref = 1.0214.
comparisons in Subsection V-C. Note that during the transients, the operating points scatter
and do not exactly follow the V-I characteristic. For a certain
|V |, the measured |Ist | and the read value from the V-
A. STATCOM models comparison and validation using real I characteristic can deviate up to 0.04 p.u. This implies
PMU data that the static STATCOM model, which is based on the V-
The real PMU data used in this subsection was recorded I characteristic, would result in deviations during transients.
during a generator trip event. The generator loss resulted in On the other hand, the pseudo-dynamic model can be a
reducing the active power flow on the main transfer paths in good choice to reflect the STATCOM’s control process and its
a neighboring system, and caused an increase in bus voltages. system dynamics. Using the transfer function estimator tfest
32.1981
A STATCOM installed on the transfer paths reacted to the in MATLAB, the transfer function is estimated as 0.0329s+1 ,
voltage change. The STATCOM bus voltage phasor, as well with two control parameters K = 32.1981 and T = 0.0329,

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

0.4
which are those in Eqs. (7)-(9). |Ist|−m
|Ist|−m

|Ist|(p.u.)
|Ist|(p.u.)
0.3 |Ist|−est
|Ist|−est
The gain K theoretically is the reciprocal of the V-I charac- 0.2
0.2
teristics’ slope. For the transient oscillation period, the slope 0.1 0.1
1.5 2 2.5
calculated by the linear regression is Xs = 0.0285, whose time (s)
1.5 2
time (s)
2.5

inverse (32.1543) is quite close to the calculated K (32.1981). 10

0
10
−10

Error(p.u.)

Error(p.u.)
which verifies that of the transfer function. In addition, the 10
−10
10
−15

time constant T is in accordance with the statement of

−20 |Ist|−residual−error −20 |Ist|−residual−error
STATCOM in [26], which says “typically about 10-50 ms 10
1.5 2 2.5
10
1.5 2 2.5
time (s) time (s)
depending on the var generator transport lag”.
In order to intuitively compare the STATCOM’s static model (a) |Ist | Static (b) |Ist | Pseudo-dynamic
with its pseudo-dynamic model, PMU data of the voltage mag- Fig. 8: Static and Pseudo-dynamic PMU-based SEs of the
nitude |V | is used as the arbitrary inputs of the static model modified WSCC 3-machine 9-bus test system
(see Eq. (5)) and the pseudo-dynamic model (see Eqs. (7)-
(9)), respectively. The static model applies the STATCOM’s
linear V-I characteristic and the pseudo-dynamic model applies and [30], respectively. In the modified WSCC 3-machine 9-bus
its first-order control model whose parameters were obtained test system, a STATCOM is installed at Bus 8; in the modified
above through model identification. These models’ outputs are KTH-Nordic 32 test system, the same STATCOM is installed
then compared to the |Ist | PMU measurement in Fig. 7. at Bus 43. Its two important parameters are K = 25, T = 0.04.
Synthetic measurements for off-line SE computations were
0.6 obtained by running time-domain simulations in the Power
PMU measurement
0.5 pseudo−dynamic model output System Analysis Toolbox (PSAT) [31], and 20 ms is selected
static model output
0.4
as the step-size to imitate the PMU data rate. For these tests,
no measurement noise was added because we aim to focus on
|Ist| [pu]

0.3
the influences of two network models on the SE accuracy. All
0.2 the weightings of the network equations and measurements
0.1 are assumed to be 1, and full measurement observability is
0
assumed.
1) Test on the modified WSCC 3-machine 9-bus system: A
−0.1
69.5 70 70.5 71 71.5 72
time [s]
72.5 73 73.5 74 16.67% load increase (both active power and reactive power)
at Bus 8 was applied at t = 2s. As shown in Fig. 8, for both
Fig. 7: PMU data, static model output and pseudo-dynamic cases the current magnitude residuals before the perturbation
model output for the current at the STATCOM occurred are below 10−13 p.u.; however, after the instance
when the perturbation occurred, the static PMU-based SE
Figure 7 compares results: the pseudo-dynamic model’ out- residuals increase up to 0.1783 p.u. then drop to 10−3 p.u.
put |Ist | coincides with the PMU data; while the static model’s while the pseudo-dynamic PMU-based SE successfully stays
output gradually deviates from the the PMU data. The reason on the same level during transient dynamics. More estimation
that the deviation is not prominent is because the time constant accuracy performances are shown in Table II.
of the STATCOM is quite close to the PMU sampling rate, 2) Test on the modified KTH-Nordic 32 system: A 33%
thereby its dynamical trajecory can be nearly tracked by the load increase (both active power and reactive power) at Bus
PMU measurement. For a larger and more complex system, 43 was applied at t = 2s. As shown in Fig. 9, the static
this deviation could be amplified further. An analysis of their SE and pseudo-dynamic SE hold similar SE accuracy of the
residuals w.r.t. the PMU data is shown in Table I. current magnitude before the perturbation. However, after the
TABLE I: STATCOM models accuracy comparison perturbation occurs, the current magnitude estimated by the
static SE shows a residual up to 0.3666 p.u. On the other
SE methods µ* δ ** Max. residual hand, the pseudo-dynamic SE gives a maximum residual of
Pseudo-dynamic 0.0012 0.0056 0.0152
5.0626 × 10−14 p.u. Table II summaries the results of these
tests.
Static 0.0198 0.0085 0.0406
* µ denotes the average value of the residuals w.r.t. the PMU data

** δ denotes the standard deviation of the residuals w.r.t. the PMU data C. VSC-HVDC model
There are many VSC-HVDC simulation models proposed in
the literature. However, in order to make the test system acces-
sible by other researchers, the VSC-Based HVDC Transmission
B. STATCOM model in two test systems Link model provided by MATLAB R2013 B/SimPowerSystems
Next, in order to further study the proposed STATCOM is used to generate synthetic measurements to validate the
model and PMU-based SE algorithm, a modified WSCC 3- proposed VSC-HVDC model and PMU-based SE algorithm.
machine 9-bus test system and a modified KTH-Nordic 32 test A detailed description of the model and control strategy can be
system are used. Their one-line diagrams can be found in [16] found in [32]. All the control parameters preset in the Simulink

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

1.5 1.5
Ivmag−true
0.6 Ivmag−true
|Ist|(p.u.)

0.6

|Iv|(p.u.)
Ivmag−est.

|Iv|(p.u.)
|Ist|(p.u.)
1 1 Ivmag−est.
0.4 0.4
0.2 0.2 0.5 0.5
|Ist|−m |Ist|−m
0 |Ist|−est 0 |Ist|−est 0 0
−0.2 −0.2 1 1.5 2 2.5 3 1 1.5 2 2.5 3
1.5 2 2.5 1.5 2 2.5 Time (s) Time (s)
Time (s) Time (s) −4
0.1 x 10
2
0.4 10
−10
Ivmag−estimation−residual Ivmag−estimation−residual

Error(p.u.)

|Iv|(p.u.)
|Ist|−residual−error 1
Error(p.u.)

Error(p.u.)
0.05
0.2 −15 0
10

|Ist|−residual−error 0 −1
1 1.5 2 2.5 3 1 1.5 2 2.5 3
0 10
−20
Time (s) Time (s)
1.5 2 2.5 1.5 2 2.5
Time (s) Time (s)
(a) |Iv | Static (b) |Iv | Pseudo-dynamic
(a) |Ist | Static (b) |Ist | Pseudo-dynamic 2
Idc−true
1.5
Idc−true

Idc(p.u.)
Idc−est.

Idc(p.u.)
1 1 Idc−est.

Fig. 9: Static and Pseudo-dynamic PMU-based SEs of the 0 0.5

modified KTH-Nordic 32 test system −1 0

1 1.5 2 2.5 3
1 1.5 2 2.5 3
Time (s) Time (s)
−7
0.2 x 10
4
1.1 1.15 Idc−estimation−residual Idc−estimation−residual

Error(p.u.)

Idc(p.u.)
Vvmag−true 2
|Vv|(pu)

|Vv|(pu)

1.1 Vvmag−est.
0.1
1.05 Vvmag−true 0
Vvmag−est.
1.05
0 −2
1 1 1 1.5 2 2.5 3 1 1.5 2 2.5 3
1 1.5 2 2.5 3 1 1.5 2 2.5 3 Time (s)
Time (s) Time (s)
Time (s)
−8
0.03
Vvmag−estimation−residual
5
x 10
Vvmag−estimation−residual
(c) Idc Static (d) Idc Pseudo-dynamic
Error(pu)

Error(pu)

0.02 0

0.01 −5 Fig. 11: Static and Pseudo-dynamic PMU-based SEs of the

0
1 1.5 2 2.5 3
−10
1 1.5 2 2.5 3 VSC-HVDC link test system’s second test scenario
Time (s) Time (s)

(a) |Vv | Static (b) |Vv | Pseudo-dynamic

2.6
Vdc−true
2.6
Vdc−true accuracy during normal operation (order of 10−15 ) while in
Vdc(p.u.)
Vdc(p.u.)

Vdc−est. Vdc−est.
2.5 2.5

2.4 2.4
Figs. 10 and 11 improvement during normal operation is seen
2.3 2.3
1 1.5 2 2.5 3
for the pseudo dynamic model. The reason is the pseudo-
1 1.5 2 2.5 3
Time (s)
−7
x 10
Time (s)
dynamic model is capable to include internal states that are
0.015 5
Vdc−estimation−residual Vdc−estimation−residual calculated in real-time, such as the voltage reference used in
Error(p.u.)

Vdc(p.u.)

0.01

0.005
0 Section IV. In contrast, static models do not possess such
0 −5
1 1.5 2 2.5 3
flexibility.
1 1.5 2 2.5 3
Time (s) Time (s)

(c) Vdc Static (d) Vdc Pseudo-dynamic VI. SE COMPUTATION PERFORMANCE

Fig. 10: Static and Pseudo-dynamic PMU-based SEs of the One advantage of the proposed pseudo-dynamic PMU-based
VSC-HVDC link test system’s first test scenario SE is that it does not significantly increase the computation
complexity and burden when compared to the static SE. By
comparing the number of iterations and computation time
model are kept the same in the pseudo-dynamic SE model. The of the static and the pseudo-dynamic PMU-based SEs, as
measurements generated by simulating the VSC-Based HVDC shown in Table III, it can be seen that the pseudo-dynamic SE
Link model are then re-sampled with 20 ms rate to imitate the performs similarly to the static SE in terms of computation
PMU data rate. speed. These SE computations were carried out on the VSC-
1) First test scenario: The inverter’s DC voltage reference HVDC test scenario I using an ordinary PC with an Intel(R)
dropped from 1 p.u. to 0.95 p.u. at t = 2.1s. As shown in Core(TM) i7-2640M CPU @2.80GHz and a 8.00 GB RAM,
Fig. 10, the pseudo-dynamic SE performs more accurately by and using MATLAB R2013 B .
a factor of 1/1000 than the static SE not only during transients,
but also during steady state. For instance, the voltage magni- TABLE III: Computation performance comparison
tudes on the rectifier side estimated by the static SE shows PMU-based Aver. comp. time Aver. no. of iter. Largest no.
a maximum residual up to 0.0279 p.u. while the pseudo- SE method per snapshot per snapshot of iteration
dynamic SE gives a maximum residual of 7.6517 × 10−8 p.u.
P-dynamic 4.754 ms 5.465 10
2) Second test scenario: A three-phase line breaker on the
Static 3.115 ms 5.525 11
inverter side was opened from t = 2.1s for 0.12s. For this
larger perturbation, as shown in Fig. 11, the pseudo-dynamic
SE performs more accurately by a factor of 1/1000 than the
static SE. For instance, the DC current estimated by the static VII. C ONCLUSION
SE shows a maximum residual up to 0.1418 p.u. while the This article presents a PMU-based state estimator using
pseudo-dynamic SE gives a maximum residual of 2.9504 × a pseudo-dynamic network model. This method significantly
10−7 p.u. Table II summaries the results of these tests. improves the SE accuracy during transients as compared to
As it might be noted that in Figs. 8 and 9 both static and the static PMU-based SE. In contrast to most dynamic SE
pseudo dynamic models maintain almost the same level of algorithms, it implements an iterative algorithm to update the

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

TABLE II: SE accuracy performance on test systems

Before perturbation (µ, σ) After perturbation (µ, σ) Max. residual
static pseudo-dynamic static pseudo-dynamic static pseudo-dynamic

Res.|V8 | (1.57e-16, 1.02e-16) (1.57e-16, 1.02e-16) (3.85e-16, 1.27e-15) (7.25e-16, 1.76e-15) 7.11e-15 8.22e-15
STATCOM test1
Res.|Ist | (-3.84e-15, 9.58e-15) (-3.72e-15, 2.01e-15) (8.03e-04, 0.0127) (-3.57e-15, 2.16e-14) 0.1783 1.05e-13

Res.|V43 | (-8.71e-17, 7.13e-17) (-8.71e-17, 7.13e-17) (-2.94e-17, 2.12e-16) (-2.39e-17, 1.95e-16) 2.44e-15 2.22e-15
STATCOM test2
Res.|Ist | (-2.06e-16, 6.37e-15) (2.11e-15, 1.83e-15) (0.0048, 0.0332) (1.57e-15, 7.16e-15) 0.3666 5.06e-14

Res.|Vv | (0.0023, 0.0024) (5.45e-10, 4.74e-09) (0.0187, 0.0039) (-5.39e-08, 1.78e-08) 0.0279 7.65e-08
Res.|Iv | (-0.0091, 0.0035) (-1.59e-08, 4.37e-07) (-0.0135, 0.0042) (5.60e-08, 2.36e-06) 0.0214 1.04e-05
VSC-HVDC test1
Res.Vdc (-0.0012, 0.0010) (-6.60e-10, 1.09e-07) (-0.0094, 0.0021) (1.05e-08, 1.25e-07) 0.0141 4.44e-07
Res.Idc (0.0108, 0.0041) (-3.46e-11, 2.00e-09) (0.0148, 0.0045) (2.00e-08, 7.62e-09) 0.0237 2.81e-08

Res. |Vv | (0.0023, 0.0024) (5.45e-10, 4.74e-09) (-0.0131, 0.0423) (-1.62e-08, 3.77e-08) 0.0207 1.87e-07
Res. |Iv | (-0.0091, 0.0035) (-1.59e-08, 4.37e-07) (-0.0059, 0.0174) (1.18e-06, 1.84e-05) 0.0824 8.15e-05
VSC-HVDC test2
Res.Vdc (-0.0012, 0.0010) (-6.60e-10, 1.09e-07) (0.0026, 0.0119) (9.57e-07, 9.75e-07) 0.0400 3.44e-06
Res. Idc (0.0108, 0.0041) (-3.46e-11, 2.00e-09) (0.0068, 0.0307) (1.01e-08, 5.23e-08) 0.1418 2.95e-07

estimated states instead of solving differential and algebraic [8] M. Filho and J. Souza, “Forecasting-Aided State Estimation—Part I:
equations (DAEs), which significantly saves computational Panorama”, IEEE Trans. Power Syst., vol. 24, no. 3, Nov. 2009.
[9] H. R. Sirisena and E. P. M. Brown, “Composite Hydroelectric Energy
resources. Additionally, the pseudo-dynamic network model System State Estimation”, IFAC Symposia Series, Pergamon, Oxford,
can be easily constructed from the original static model and 1987, Pages 399-404, Power Systems and Power Plant Control.
represent devices with time-varying control references. [10] B. Xu, and A. Abur , “State Estimation of Systems with Embedded
FACTS devices,” IEEE PowerTech Conference, Jun. 2003.
The second major contribution of this article is the pseudo- [11] C. Rakpenthai, S. Premrudeepreechacharn, S. Uatrongjit, “Power system
dynamic network models for STATCOMs and VSC-HVDCs. with multi-type FACTS devices state estimation based on predictor-
These two models also provide valuable and practical insight corrector interior point algorithm”, Electr. Power Energy Syst., vol. 31,
no. 4, pp. 160-166, May 2009.
on how to develop pseudo-dynamic network models for com-
[12] A. Zamora-Cárdenas, C. R. Fuerte-Esquivel, “State estimation of power
ponents and controllers in power systems. systems containing FACTS controllers”, Electr. Power Energy Syst., vol.
The case studies provide sufficient evidence that the pseudo- 81, no. 4, pp. 995-1002, April 2011.
dynamic SE algorithm is capable of performing much more [13] P. I. Bartolomey, S. A. Eroshenko, E. M. Lebedev, and A. A. Suvorov,
“New information technologies for state estimation of power systems with
accurate estimation during transient dynamics without signif- FACTS,” IEEE PES ISGT Europe, Oct. 2012.
icantly increasing the computation resources as compared to [14] E. A. Zamora-Crdenas, B. A. Alcaide-Moreno, C. R. Fuerte-Esquivel,
the static SEs. The STATCOM using real PMU data shows that “State estimation of flexible AC transmission systems considering syn-
chronized phasor measurements,” Electr. Power Syst. Res., vol. 106, pp.
the proposed modeling approach and PMU-based algorithm is 120-133, January 2014.
applicable when real PMU data from actual power systems is [15] V. I. Presada, C. V. Cristea, M. Eremia, and L. Toma, “State estimation
available. in power systems with FACTS devices and PMU measurements,” UPEC,
Sept. 2014.
Future work will focus on testing the proposed approach [16] W. Li, and L. Vanfretti, “A PMU-Based State Estimator for Networks
under noisy measurements w.r.t. the computational burden of Containing FACTS Devices,” IEEE PES PowerTech, June 2015.
the SE. [17] H. R. Sirisena, E. P. M. Brown, “Inclusion of HVDC links in AC power-
system state estimation,” , IEE Proc. Gen. Trans. Dist., vol. 128, no. 3,
pp. 147-154, May 1981.
R EFERENCES [18] Q. Ding, B. Zhang, and T. S. Chung, “State estimation for power systems
[1] S. G. Ghiocel et al., “Phasor-Measurement-Based State Estimation for embedded with FACTS devices and MTDC systems by a sequential
Synchrophasor Data Quality Improvement and Power Transfer Interface solution approach”, Electr. Power Syst. Res., vol. 55, no. 3, pp. 147-156,
Monitoring,” IEEE Trans. Power Syst., vol. 29, no. 2, pp. 881-888, March Sept. 2000.
2014. [19] W. Li, and L. Vanfretti, “A PMU-Based State Estimator considering
[2] E. R. Fernandes et al., “Application of a Phasor-Only State Estimator to classic HVDC links under different control modes,” Sustainable Energy,
a Large Power System using Real PMU data,” IEEE Trans. Power Syst., Grids and Networks, vol.2, pp. 69-82, 2015.
vol.PP, no.99, pp.1-1, March 2016. [20] A. V. Jaén, E. Acha, and A. G. Expósito, “Voltage Source Converter
[3] L. Vanfretti, J. H. Chow, S. Sarawgi, B. Fardanesh, “A Phasor-Data-Based Modeling for Power System State Estimation: STATCOM and VSC-
State Estimator Incorporating Phase Bias Correction,” IEEE Trans. Power HVDC”, IEEE Trans. Power Syst., vol. 23, no. 4, Nov. 2008.
Syst., vol. 26, no. 1, pp. 111-119, Feb. 2011. [21] J. Cao, W. Du, H .F. Wang, “The incorporation of generalized VSC
[4] A. G. Phadke, and J. S. Thorp, Synchronized phasor measurements and MTDC model in AC/DC power system state estimation,” SUPERGEN
their applications. New York: Springer Science & Business Media, 2008. Sept. 2012.
[5] N. Zhou and Z. Huang, et al. (2014 Mar.). Capturing Dynamics in the [22] E. A. Zamora-Crdenas, C. R. Fuerte-Esquivel, A. Pizano-Martnez,
Power Grid: Formulation of Dynamic State Estimation through Data H. J. Estrada-Garca, “Hybrid state estimator considering SCADA and
Assimilation. PNNL, WA. [Online]. Available: http://www.pnnl.gov/main/ synchronized phasor measurements in VSC-HVDC transmission links”,
publications/external/technical reports/PNNL-23213.pdf Electr. Power Syst. Res., vol. 133, pp. 42-50, April 2016.
[6] F. C. Schweppe and R. D. Masiello, “A Tracking Static State Estimator”, [23] W. Li, and L. Vanfretti, “A PMU-Based State Estimation For Networks
IEEE Trans. Power App. Syst, vol. PAS-90, no. 3, pp. 1025-1033, May Containing VSC-HVDC links,” IEEE PES General Meeting, July 2015.
1971. [24] P. Yang, Z. Tan, A. Wiesel and A. Nehorai, “Placement of PMUs
[7] A. S. Debs and R. E. Larson, “A Dynamic Estimator for Tracking the Considering Measurement Phase-Angle Mismatch,” IEEE Trans. Power
State of a Power System”, IEEE Trans. Power App. Syst, vol. PAS-89, Delivery, vol. 30, no. 2, pp. 914-922, April 2015.
no. 7, pp. 1670-1678, Sept. 1970. [25] M. S. Almas, L. Vanfretti, R. S. Singh, and G. M. Jonsdottir, “Vulnera-

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2777185, IEEE Access

10

bility of Synchrophasor-based WAMPAC Applications to Time Synchro-

nization Spoofing,” IEEE Trans. Smart Grid, vol.PP, no.99, pp.1-1.
[26] N. G. Hingorani, and L. Gyugyi, “Chapter 5 Static Shunt Compensators:
SVC and STATCOM,” in Understanding FACTS–Concepts and Technol-
ogy of Flexible AC Transmission Systems. New York: IEEE Press, 2000.
[27] K. K. C. Yu and N. R. Watson, “An Approximate Method for Transient
State Estimation,” IEEE Trans. Power Delivery, vol. 22, no. 3, pp. 1680-
1687, July 2007.
[28] L. Harnefors and H. -P. Nee, “Model-based current control of AC
machines using the internal model control method,” IEEE Trans. Industry
App., vol. 34, no. 1, pp. 133141, 1998.
[29] S. Cole, J. Beerten and R. Belmans, “Generalized dynamic VSC MTDC
model for power system stability studies,”IEEE Trans. Power Syst., vol.
25, no. 3, pp. 16551662, Aug. 2010.
[30] Y. Chompoobutrgool, W. Li, and L. Vanfretti, “Development and Imple-
mentation of a Nordic Grid Model for Power System Small-Signal and
Transient Stability Studies in a Free and Open Source Software”, IEEE
PES General Meeting, July 2012.
[31] F. Milano, “An Open Source Power System Analysis Toolbox,” IEEE
Trans. Power Syst., vol. 20, no. 3, pp. 1199-1206, Aug. 2005.
[32] VSC-Based HVDC Link, R2015b, MathWorks, Inc., Natick, MA. [On-
line]. Available: http://se.mathworks.com/help/physmod/sps/powersys/ug/
vsc-based-hvdc-link.html

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.