Smart Distribution Transformer for Secondary Voltage Regulation

and Load Modeling

Josemar O. Quevedo, Rafael C. Beltrame, Mauricio Sperandio,

Tiago Marchesan, Luciano Schuch
FEDERAL UNIVERSITY OF SANTA MARIA – UFSM
POWER ELECTRONICS AND CONTROL RESEARCH GROUP – GEPOC
97105-900, Santa Maria, RS, Brazil
Tel.: +55 / (55) – 3220.9497. Fax: +55 / (55) – 3220.9497.
E-Mail: [email protected]
URL: http://www.ufsm.br/gepoc

Acknowledgements
The authors would like to express their gratitude to “Conselho Nacional de Desenvolvimento
Científico e Tecnológico (CNPq),” “Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
(CAPES)” and the “Centrais Hidrelétricas de Carazinho” (Eletrocar) through the P&D ANEEL
Project, for financial support.

Keywords
«Distribution of electrical energy», «energy system management», «power quality», «transformer»,
electronic on-load tap changer, load identification.

Abstract
The distribution networks have been subjected to increased demand for electricity and the inclusion of
distributed generation (DG), which have affected the voltage regulation. In this work, it is proposed a
smart distribution transformer, which encompasses an electronic on-load tap changer and a
communication system. In addition to the automatic or remote voltage regulation, the system enables
to identify the load model at the interface between primary and secondary distribution systems. These
characteristics permit the enhancement of the power quality, along with the improvement of the
distribution system operation, planning and revenue management. The main focus of this work is the
load parameter identification, which is validated by a simulation tool and experimental results.

Introduction
The electricity distribution networks were originally designed for radial power flow, but nowadays,
they have increasingly received injection of energy from distributed generation (DG) sources. Allied
to this, the growth in demand for electricity associated with more strict power quality standards have
increased the distribution systems complexity, complicating the operation, planning and service
quality in these systems.

One of the most affected parameters due the injection of DG is the voltage level delivered to the end
users. This is caused by the change in both the magnitude and direction of power flow on the network,
being a result of generation variability, a common characteristic of renewable sources such as
photovoltaic and wind power [1]. Thus, strategies that allow the grid voltage regulation in the presence
of DG are fundamental.

Another factor that directly impacts on the voltage levels is the increasing demand for electricity. Due to
this, there is a tendency for the power systems to be operated close to their limits, requiring load
shedding in overloaded situations [2]. However, the actions for reducing the load usually occur in broad
regions and may represent losses for users and power utility companies. In these cases, the optimization

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 of the load shedding through localized actions can reduce the damage caused. Nevertheless, for
improving the load reduction, it is necessary appropriate load models and equipment that allow power
flow control along the network, enabling a better planning for the system operation [3].

The most popular method for voltage regulation in distribution systems is the employment of
transformers with tap changers. Even though no-load tap changers present a low cost for
implementation and high reliability, they cannot perform the automatic voltage regulation, requiring
human intervention in the tap commutation process. On the other hand, transformers equipped with
electromechanical on-load tap changers (OLTC) allow the automatic voltage regulation, and,
consequently, the electricity demand control. However, due to high implementation and maintenance
costs, associated with the slow commutation of the electromechanical switches, these devices are
mainly used in substation transformers and as primary network voltage regulators.

An alternative that has been envisioned for regulating the voltage delivered to consumers is the use of
electronic tap changers, which allow a faster response, commutation without electric arc and
additional features to distribution transformers, such as the automatic voltage control at the interface
between primary and secondary distribution grid [4]. This feature is even more promising when
combined with a suitable communication system. In this case, it is possible to obtain remote control
over the equipment, as well as, the measurement of variables at the point of common coupling (PCC),
network management, among others.

In this work, a distribution transformer equipped with an electronic OLTC employing a communication
system is incorporated with the load parameter identification function. This functionality is based on the
ZIP model and allows identifying the proportion of load types connected to the transformer, i.e., constant
impedance, constant current and constant power. The proposed system permits the adjustment of the
voltage automatically or remotely assisted, enabling to set the voltage level between standardized limits.
In addition, based on the obtained load model, it is possible to actuate over the transformer taps in order
to obtain greater control for revenue and load shedding close to the end users.

Smart distribution transformer

The smart transformer which employs an electronic OLTC is shown in Fig. 1. It is used a 5 kVA
single-phase earth return (SWER) distribution transformer, with the primary voltage rates defined by
7967/7621/7274/6928/6581 V, and secondary voltages defined by 440/220 V. The system employs
Insulated Gate Bipolar Transistors (IGBTs) and diodes in order to obtain bidirectional switches. The
activation of the electronic switches is via optical channels, which ensure high electrical insulation and
noise immunity. From Fig. 1 (a), it can be noticed the presence of a current transformer (CT) in the
primary circuit, which is used to enable the commutation of taps during the zero crossing of the
current, it is also employed a potential transformer (PT) which feeds the gate-driver circuits,
instrumentation, control and the communication systems. The control system is composed of a
TMS320F28335 Texas Instruments® DSP and interface circuits.

The system also measures the secondary voltage, for regulating the voltage delivered to end users, and the
secondary current, in order to allow the calculation of active, reactive and apparent power. It should be
noted that the rms values of the measured variables: primary and secondary voltages, as well as, primary
and secondary currents, are transmitted periodically to the power utility through the communication
system. It is also transmitted the transformer oil temperature, the tap position, the transformer operation
mode (automatic or management) and the load parameters based on the ZIP model.

By employing a bidirectional communication system with the power utility company, it is possible to
actuate remotely on the transformer. The direct operating actions made possible by this system are
defined as follows: (i) definition of transformer operation mode for voltage regulation (automatic or
remote management); (ii) remote definition of which tap the transformer must operate; (iii) load
shedding and (iv) revenue control through the variation of the supplied voltage levels.

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 (a) (b)
Fig. 1: Smart transformer. (a) Schematic. (b) Developed system.

In addition, by processing the data obtained from the transformer, it is possible to execute the following
indirect actions for network management: (i) planning of maintenance, expansion and network
modifications; (ii) verification of the energy demanded in the transformer secondary side, allowing to
identify inconsistencies between provided and billed energy, whether caused by power loss or theft; (iii)
reactive correction necessity; (iv) identification of overload; (v) load profile and consumption pattern of
users; (vi) power flow studies and feasibility for inclusion of DG; among others.

Load modeling
For planning and operating the electrical systems cost effectively and safely it is important to set a
suitable model for each system element. Various studies have been conducted in order to improve the
models of the system components, such as: transmission and distribution lines, transformers,
generators, etc. Using appropriate models various grid aspects can be analyzed, as fault effects,
maintenance scheduling and future connections of users. These models are even more necessary in
intelligent networks, where information about the power grid is essential for its proper operation [5].

Even though it represents one of the most important aspects for the power grid management, the load
models have shown the greater difficulty in the system representation [6]. One of the main difficulties
in modeling the load is that it changes continuously, reflecting the user consumption patterns that
varies due to the time of day, weather conditions, state of the economy, among other factors [7]-[8].
For this reason, although it is possible to know how each load element behaves individually, it is not
possible to have a precise notion of the final composition.

Some key factors that influence the difficulty for obtaining a load model are [9]: (i) the large number and
diversity of equipment; (ii) location of the property and inaccessibility of the load devices for the power
utility; (iii) change of the load composition during the day, week, season, weather, and time of day; (iv)
uncertainties about the load components, especially for large variations in voltage or frequency.

The load models are usually classified into two broad categories: dynamic models and static models
[8], [10]. For obtaining the load models there are two main approaches: one based on the components
and other based on measurements [5], [8]-[9]. Conventionally, the approach based on measurements is
the most used due to the uncertainties involved in the component based method. In this work, the
measurement method is used to define the load static model.
Load static models
In the studies of network planning and analysis, static models are more conventionally used. By
definition, the load characteristic is considered static when it can be represented by polynomials or

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 other algebraic equations, such as exponential functions [5], [7], [11]. In these cases, it is common to
represent the loads by separate consideration of active and reactive power [7].

Traditionally, load models reflect the sensitivity of the power with respect to voltage variations. This
dependency characteristic of the load voltage has been represented as an exponential model [8], [11]
and can be represented by (1) and (2):

a
P Po V (1)
b
Q Qo V (2)

where Po and Qo represent the initial values of active and reactive power of the system. When
superscript a or b are 0, 1 or 2, they represent, respectively, load characteristic of constant power,
constant current or constant impedance. An alternative model that has been widely used is the
polynomial model [8], given by (3) and (4):

P Po p1  p2V  p3V 2 (3)

Q Qo q1  q2V  q3V 2 (4)

This model is commonly known as ZIP model, i.e., constant impedance (Z), constant current (I) and
constant power (P). The coefficients p1 to p3, q1 to q3 define the proportions of each component of the
model, in which p1 and q1 represent the components of constant power, p2 and q2 represent constant
current and p3 and q3 represent constant impedance.

In this work, the model used to describe the load is the ZIP model. This is determined from changes in
the transformer secondary voltage through the taps commutation. For each operating condition, the
rms values of voltage, the active power and reactive power are measured and stored, and subsequently,
they are used to obtain the coefficients that describe the ZIP model.
ZIP model coefficients definition
The definition of the ZIP model coefficients can be performed through a curve fitting method. In this
paper, the method used is the Ordinary Least Squares (OLS) [13]. The definition of the samples used
in the OLS method is performed as follows:

x Set up an initial condition for operation of the electronic OLTC and to perform the
measurement of voltage, active power and reactive power when the transformer is in static
operation. In the proposed system, the defined starting position is the one which defines the
lowest value that attends the standard operation tolerance for secondary voltage.
x In the following, the electronic OLTC is incremented in the tap position, so it can be obtained
static values of voltage and power for each different configuration of the electronic OLTC. In
the proposed system, these variations occur starting from the initial condition until the
maximum voltage value that still attends the standard definition for the operating tolerance of
the secondary voltage.
x Once obtained all the necessary voltages, active and reactive power values, it can be obtained
the ZIP model load parameters from the adjustment of the OLS method.
x After the identification of load model parameters, the transformer can return to the initial
operating mode, either automatic or managed by the power utility.
Ordinary Least Squares Method
The ZIP model coefficients for active power and reactive power, given respectively given by (3) and
(4), are obtained through the OLS method. This method consists of finding a function that fits a set of
samples so as to minimize the sum of squared errors resulting from the measured values and the
adjusted curve.

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 Considering (5), yi (i = 1...p) represent the active power and reactive power measurements, xi are the
voltage measurements, İi are the residues between the measured values and the adjusted curve,
ȕk (k = 0, 1, 2...n) are the coefficients for the polynomial curve, p is the number of measurements for
voltage and power and n is the polynomial order.

yi E0  E1 xi  E 2 xi 2  ...  E n xi n  H i (5)

The curve adjustment for the samples xi e yi can be obtained by minimizing the residues İi:

Hi yi  E 0  E1 xi  E 2 xi 2  ...  E n xi n (6)

Rewriting (5) in a matrix form:

ª y1 º ª1 x1 x12 ! x1n º ª E 0 º ªH1 º

«y » « » « » « »
« 2» «1 x2 x2 2 ! x2 n » « E1 » «H 2 »
«# ˜  (7)
«# » # # ! # » «# » «# »
« » « » « » « »
«¬ y p »¼ «¬1 E n ¼ «¬H p »¼
x p x p 2 ! x p n »¼ ¬N
N  N
Y B E
X

It is possible to demonstrate that the solution for the coefficient vector B is given by:
1
B XTX ˜ X TY (8)

where XT represents the transpose matrix of X. From (8), it is possible to verify that the coefficients ȕk,
that in the ZIP model are represented by p1 a p3 for the active power and q1 a q3 for reactive power,
depend exclusively from the voltage and power values measured.

Secondary voltage regulation method

The proposed method for controlling the transformer secondary voltage aims at adjusting the
secondary voltage levels within the operating tolerable values defined by standardization, such as
ANSI C84.1. For instance, in the case of Brazilian standardization PRODIST [14], for SWER
networks, the tolerable values for 220 V grids are between 202 V and 231 V, for 440 V grids the
tolerable values are between 405 V and 462 V.

Thus, by allowing the automatic adjust of the voltage, the use of the proposed device enables to improve
the power quality for end users and service indicators for power utilities. Furthermore, it should be noted
that with a regulation tolerance margin, the secondary voltage can be adjusted for different conditions
without disobeying the standardized voltage levels. This approach favors additional functionalities to the
equipment, for example, load shedding, losses and revenue optimization.

These functionalities are improved when the load model is known. For instance, when the transformer
is loaded predominantly with constant impedance load (Z), there is a quadratic reason between voltage
and power, as can be noticed in Fig. 2. In these cases, by increasing the system voltage within the
tolerable values, the billed power is increased quadratically. The opposite is valid for overloaded
systems, where voltage reduction allows load shedding by a quadratic reason. On the other hand,
transformers loaded mainly with constant current load (I) allow the system power to vary
proportionally to the voltage, without changing the current and losses of the network, as can be seen in
Fig. 2. Also, transformers with a large proportion of constant power load (P) allow the system losses
reduction with the increase in voltage, but prevent the load shedding or revenue management. In this
way, by knowing the load behavior, it is possible to control the voltage levels in order to obtain better
management of the power grid.

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 Fig. 2: Influence of voltage variation on different load types: constant impedance (Z), constant current
(I) and constant power (P).

Simulation results
The proposed system was simulated by considering the parameters summarized in Table I. In order to
facilitate the interpretation, the system parameters are given per unit (p.u.). The load is composed by
the three different load types and each one represents 1 p.u. The constant impedance load has a 0.8
inductive power factor (PF), the constant current load has a 0.8 inductive PF and the constant power
load is purely active power. Furthermore, the lowest voltage level provided by the transformer (1 p.u.)
has been defined as the initial condition, which occurs when the tap 1 is activated, as can be noticed in
Fig. 3. When considering Fig. 1, this condition occurs when the electronic switch S1 is switched on.

In order to obtain the static load response for each voltage level, there is a delay of twenty cycles after
each tap commutation, avoiding the transitory effects of the voltage change. After this period, it is
performed the calculation of the rms secondary voltage values, active power (P) and reactive power
(Q) during two cycles of the grid frequency. The calculated values are stored for obtaining the ZIP
model coefficients. After this, a new commutation is enabled and the same procedure is performed.

In Fig. 3 it is shown the data obtained from the simulation realized through the software PSIM®.
Table II summarizes this data, indicating the rms values for secondary voltage and current, the active
power, reactive power and apparent power for the operation of different transformer taps. Table III
presents the calculated coefficients for (3) and (4), obtained through the OLS method. It should be
noted that Po and Qo values are defined by the initial condition, i.e., when the tap 1 is selected.
Table I: Simulation data
Parameter Value
Initial secondary voltage 1 p.u.
Constant Z load 0.8 + j0.6 p.u.
Constant I load 0.8 - j0.6 p.u.
Constant P load 1 p.u.
Table II: Values of voltage, current and power
Parameter Tap 1 Tap 2 Tap 3 Tap 4 Tap 5
Secondary rms voltage 1.000 p.u. 1.045 p.u. 1.095 p.u. 1.150 p.u. 1.210 p.u.
Secondary rms current 2.863 p.u. 2.868 p.u. 2.878 p.u. 2.893 p.u. 2.913 p.u.
Active power 2.600 p.u. 2.710 p.u. 2.836 p.u. 2.978 p.u. 3.141 p.u.
Reactive power 1.200 p.u. 1.283 p.u. 1.376 p.u. 1.483 p.u. 1.605 p.u.
Apparent power 2.863 p.u 2.999 p.u 3.152 p.u 3.327 p.u 3.527 p.u
Table III: ZIP model coefficients
Coefficient Coefficient
Value Value
(active power) (reactive power)
p1 0.384609 q1 -6.6903e-7
p2 0.307714 q2 0.4999545
p3 0.307676 q3 0.5000461
R2 0.99999997 R2 0.99999999

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 In Fig. 4 it is possible to compare the sampled values and the curves generated by the ZIP model for
active power (P) and reactive power (Q) as a function of the secondary voltage. In this figure it is also
shown a curve that represents the apparent power (S). The coefficient of determination (R2) presented
in Table III indicates that the measured values for active and reactive power are well described for the
respective ZIP models found.

Based on the ZIP models obtained for active and reactive power, this analysis can be extended to the
evaluation of the system current (IS), as shown in Fig. 5, which can be described in terms of a
component related to the active power (IP) and other related to the reactive power (IQ).

Regarding the studied case, from the ZIP model it is possible to identify that the largest share of the
active power is due to the constant power section, which also includes equal portions of constant
impedance and constant current. This behavior can be identified on Fig. 4, where the increase of the
voltage results in the increase of active power. Furthermore, as can be seen in Fig. 5, due to higher
share of constant power component and constant current, the current IP does not increase significantly
when the voltage rises.

Concerning the reactive power ZIP model, there is only the presence of constant current and constant
impedance components. As can be seen in Fig. 2, the voltage rise causes a significant increase in
reactive power. This effect is mainly caused by the constant impedance component due to its quadratic
relation to the voltage. Also, due to the constant impedance load, the voltage elevation causes a
proportional increase in current IQ, since the constant current component undergoes no change with the
variation of the system voltage.

Thus, it is possible to verify for the studied case that increasing the system voltage implies no
significant increase in current. Furthermore, there is a significant elevation in both active and reactive
power with the voltage raise. In this case, when the voltage levels are elevated, within the limits set by
the operation tolerance defined in standards, it would be allowed higher revenue for the power utility
without significant increase in losses. Furthermore, the voltage reduction permits load shedding, but
does not significantly reduces the system losses.

Fig. 4: Comparison between ZIP model and

sampled values for apparent power (S), active
power (P) and reactive power (Q).

Fig. 3: Load identification simulation. Superior:

electronic switches logic state (S1 a S5). Center: Fig. 5: Comparison between ZIP model and
secondary voltage. Inferior: Active power (P), sampled values for secondary current.
reactive power (Q) and apparent power (S).

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 Experimental results
The experimental results were performed in the transformer equipped with the electronic OLTC
presented in Fig. 1 (b). The measurements were validated through a digital power meter Yokogawa®
WT1600, with maximum error between measures inferior to 0.1%. Two types of load have been
evaluated: a pure resistive load, and a resistive-inductive load. The procedure for obtaining the
measurements of voltage, current and power for each transformer tap are the same presented in the
simulation results section.

The transformer is initialized in Tap 1 (switch S1 in on state) with nominal primary voltage. In Fig. 6 it
is presented the secondary voltage behavior during the commutation of taps for pure resistive load
connected to the transformer secondary side. The secondary rms voltage values for each tap are
presented in Table IV, as well as the secondary rms current, active power, reactive power and apparent
power.

Fig. 6: Transformer secondary voltage for tap commutations.

Table IV: Measurements for the pure resistive load

Parameter Tap 1 Tap 2 Tap 3 Tap 4 Tap 5
Secondary rms voltage 257.52 V 226.95 V 238.09 V 249.82 V 262.11 V
Secondary rms current 7.83 A 8.16 A 8.54 A 8.95 A 9.38 A
Active power 1704.2 W 1852.5 W 2034.3 W 2235.5 W 2458.7 W
Reactive power 20.3 VAr 23.6 VAr 26.1 VAr 30.1 VAr 35.2 VAr
Apparent power 1704.3 VA 1852.6 VA 2034.4 VA 2235.7 VA 2458.9 VA

The same procedure was performed for a resistive-inductive load. In this case, the secondary rms
voltage values for each tap are presented in Table V, as well as the secondary rms current, active
power, reactive power and apparent power.
Table V: Measurements for the resistive-inductive load
Parameter Tap 1 Tap 2 Tap 3 Tap 4 Tap 5
Secondary rms voltage 218.40 V 227.86 V 238.75 V 250.32 V 262.54 V
Secondary rms current 7.65 V 7.96 A 8.32 A 8.73 A 9.14 A
Active power 1293.8 W 1410.6 W 1545.5 W 1700.3 W 1869.6 W
Reactive power 1057.2 VAr 1142.6 VAr 1251.1 VAr 1372.6 VAr 1506.6 VAr
Apparent power 1670.8 VA 1815.4 VA 1988.5 VA 2185.2 VA 2401.3 VA

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 The ZIP model coefficients and the coefficient of determination (R2) for the pure resistive load are
presented in Table VI. The same coefficients for the resistive-inductive load are presented in
Table VII.

Table VI: ZIP model for Table VII: ZIP model for
resistive load resistive-inductive load
Coefficient Value Coefficient Value Coefficient Value Coefficient Value
p1 0.064 q1 3.203 p1 -0.019 q1 0.372
p2 -0.078 q2 -6.866 p2 0.038 q2 -0.606
p3 1.014 q3 4.677 p3 0.981 q3 1.233
R2 0.9999 R2 0.9957 R2 0.9999 R2 0.9999

The adjusted ZIP model for the pure resistive load is presented in Fig. 7. In this figure, the reactive
power of ZIP model was not presented due to the small values involved. As can be seen, the
measured active power (P) and apparent power (S) correctly describe the adjusted model, presenting
a predominantly constant impedance behavior. The same characteristic can be observed in the
secondary rms current for the measured and adjusted values, which is presented in Fig. 8.

In Fig. 9 it is presented the adjusted ZIP model for resistive-inductive load. In this case, the measured
active power (P), reactive power (Q) and apparent power (S) are well adjusted to the obtained ZIP
model, describing the predominant constant impedance behavior for voltage variations. This
characteristic can also be visualized in Fig. 10, which describes the secondary rms current behavior for
variations on the voltage level.

Fig. 7: Comparison between ZIP model and Fig. 9: Comparison between ZIP model and
measured values for apparent power (S) and active measured values for apparent power (S), active
power (P) of the resistive load. power (P) and reactive power (Q) of the resistive-
inductive load.

Fig. 8: Comparison between ZIP model and Fig. 10: Comparison between ZIP model and
measured values for secondary current of the measured values for secondary current of the
resistive load. resistive-inductive load.

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app
 Conclusion
The proposed system has been simulated and experimentally tested for different load conditions in a
5 kVA, 7.69 kV/220V single phase earth return (SWER) transformer. The obtained results confirm the
availability of the application of the proposed system for obtaining the load ZIP model directly in the
secondary side of the transformer.

In this sense, the employment of the proposed smart transformer with a proper communication system
allows automatic and assisted voltage regulation, within the transformer operating limits, directly at
the interface between primary and secondary networks. The proposed system also enables obtaining
the network load parameters where the transformer is installed. Moreover, it offers localized control
for the power provided to end users, favoring actions in the optimization of revenue, load shedding
and the grid losses. This improves the quality of power delivered to end users and the power utility
profitability.

The acquired information from the grid, through the proposed system, can also be used for expansion
planning, maintenance and operation of the electric grid. Also, a proper processing of these data
allows the identification of regions with overload problems, areas with critical voltage regulation and
obtaining the load profile of a given region at any time.

References
[1] K. Qian, C. Zhou, M. Allan, Y. Yuan. “Lo ad modeling in distributed generation planning”, in Proc. Int.
Conf. on Sustainable Power Generation and Supply, pp. 1-6, 2009.
[2] S. Arnborg, G. Andersson, D. J. Hill, I. A. Hiskens. “On influence of load modeling for undervoltage load
shedding studies”, IEEE Trans. on Power Systems, vol. 13, no. 2, May 1998.
[3] E. Handschin, Ch. Dörnemann. “Bus load modeling and forecasting”, IEEE Trans. on Power Systems, vol.
3, no. 2, maio de 1988.
[4] S. M. García, J.C.C. Rodríguez, J. A. Jardini, J. V. López, A. I. Segura, and P. M. M. Cid, “Feasibility of
electronic tap-changing stabilizers for medium voltage lines – precedents and new configurations”, IEEE
Trans. on Power Delivery, vol. 24, no. 3, pp. 1490–1503, Jul. 2009.
[5] J. Hou, Z. Xu, Z. Y. Dong. “Load modeling practice in a smart grid environment”, in Proc. Int. Conf. on
Electric utility deregulation and restructuring and power technologies (DRPT), pp. 7-13, 2011.
[6] E. Vaahedi, M. A. El-Kady, J. A. Libaque-Esaine, V. F. Carvalho. “Load models for large-scale stability
studies from end-users consumption”, IEEE Trans. On Power Systems, vol. PWRS-2, no. 4, Nov. 1987.
[7] C. Concordia, S. Ihara. “Load representation in power system stability studies”, IEEE Trans. on Power
Apparatus and Systems, vol. PAS-101, no. 4, Abr. 1982.
[8] P. Kundur. “Power system stability and control”, McGraw-Hill, p. 1176, 1994.
[9] IEEE task force on Load Representation for Dynamic Performance. “Load representation for dynamics
performance analysis”, IEEE Trans. on Power Systems, vol. 8, no. 2, Maio 1993.
[10] T. Frantz, T. Gentile, S. Ihara, N. Simons, M. Waldron. “Load Behavior observer in LILCO and RG&E
Systems”, IEEE Trans. on Power Apparatus and Systems, vol. PAS-103, no. 4, Abril 1984.
[11] L. G. Dias, M. E. El-Hawary. “OPF incorporating load models maximizing net revenue”, IEEE Trans. On
Power Systems, vol. 8, no. 1, pp. 53-59, Fev. 1993.
[12] IEEE task force on Load Representation for Dynamic Performance. “Bibliography on load models for power
flow and dynamic performance simulation”, IEEE Trans. on Power Systems, vol. 10, no. 1, Fev. 1995.
[13] F. L. Quilumba, W. J. Lee, H. Huang, D. Y. Wang, R. L. Szabados. “Load model development for next
generation appliances”, in Proc. IEEE Industry Applications Society Annual Meeting (IAS), pp. 1-7, 2011.
[14] ANEEL, “Procedimentos de Distribuição de Energia Elétrica no Sistema Elétrico Nacional - PRODIST -
Módulo 8”, p. 76, 2016. (in portuguese).

horized licensed use limited to: King Mongkuts Institute of Technology Ladkrabang provided by UniNet. Downloaded on September 11,2024 at 09:16:27 UTC from IEEE Xplore. Restrictions app