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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 1

Dynamic Estimation of Electrical Demand in

Hot Rolling Mills
G. Alonso Orcajo, Josué Rodríguez D., Pablo Ardura G., José M. Cano, Member, IEEE, Joaquín G.
Norniella, Rocío Llera T., Diego Cifrián R.

Abstract— This paper proposes a method capable of A large part of the steel production costs is due to energy
reproducing the particular operating conditions of a hot strip usage. Savings can be achieved by producing in a more
mill and predicting the evolution of the main electrical variables efficient way. Furthermore, more energy efficient production
from both the characteristics of the steel to be milled and the
processes lead to a reduction of the environmental impact.
specific features of the rolling mill. The method analyzes the load
torque and the motor speed evolution in the stands of the Therefore, it is useful to identify the energy usage of the
roughing and finishing mill drives, according to the steel to be production process and also the quality of energy.
milled. In this study three types of carbon alloy steel are Energy is considered to be strategic for the iron and steel-
considered, thus involving dissimilar hardness characteristics. making sectors. Energy typically represents 15-20% of the
The main stands of the mill, the power network and the filter operating costs of a steel plant, and has also direct
banks have been modeled. The relationship between the grade of implications for greenhouse gas regulatory compliance costs.
steel and both the electrical demand and various power quality
parameters is discussed. The results can be used as a part of an Hot rolling operation represents approximately 8% of the total
expert system for the automatic estimation of the electrical energy consumption in an integrated steelworks. Hot rolling is
demand in a hot rolling mill. the process with the largest electrical energy demand,
accounting for 20% of the total consumption (approximately
Index Terms— Electrical demand, finishing mill, hot rolling 80 kWh/ton), and one of the most critical ones in terms of
mill, power system harmonics, roughing mill, steel. power quality. Not only is the average load considerable, but
the dynamic load variation is also large because the mill
repeatedly runs through its process cycle [1].
I. INTRODUCTION
The aim of this paper is to obtain a better understanding of

S TEELMAKING is an energy-intensive process. Although

the majority of energy is consumed by the upstream
electric energy usage in the hot rolling operation and to define
and validate major opportunities for daily energy predictions.
processes (e.g. blast furnaces, basic oxygen furnaces and
A. Short-time prediction of electrical demand and power
electrical arc furnaces), the energy consumption in the
quality parameters
downstream mills is far from insignificant. Out of the
downstream processes, the hot rolling operation is certainly It is crucial to estimate the power demand of a steel plant
the largest consumer of energy, both in the form of fuel gas both in the design stage and under different operating
and electricity. conditions. Estimation allows for guaranteeing the ability of the
The electrical consumption in the hot rolling operation is plant to withstand the anticipated load, calculating the nominal
more than 70 kWh/ton. The main consumers are the rolling capacity of the lines, designing the protections, etc. Hybrid
stands and the coilers. However, auxiliary equipment cannot solutions for filtering and compensating for reactive power are
be neglected because it represents 25% of the electrical energy becoming more common in this type of facilities. Such an ad
[1], [2]. hoc design requires a thorough analysis of reactive power
variations and a study of the planned injection of harmonics into
the network. Moreover, the aforementioned estimation is
Paper 2015-METC-0373, presented at the 2015 IEEE Industry necessary to ensure the stability of the voltage at the point of
Applications Society Annual Meeting, Dallas, Texas, October 19-22, and common coupling, the electromagnetic compatibility of the
approved for publication in the IEEE TRANSACTIONS ON INDUSTRY system with the network, the minimization of operation losses
APPLICATIONS by the Metals Committee of the IEEE Industry Applications
Society. Manuscript submitted for review July 15, 2015 and released for and the elimination of penalties from utilities [3].
publication February 3, 2016. This work was supported in part by the Most of the electrical variables that characterize power
Research and Development Center of ArcelorMittal (Avilés, Spain), and the
Ministry of Economy and Competitiveness of Spain (project within the
quality and the electromagnetic compatibility of a hot rolling
framework of the National Plan of Research, Development and Innovation, mill are related to the particular rolling conditions at the instant
reference ENE2014-52272-R). of analysis. Considering the high order of magnitude of the
G. Alonso Orcajo, Josué Rodríguez D., José M. Cano and J. G. Norniella driven power and the strongly non-linear characteristic of the
are with the Electrical Engineering Department, University of Oviedo, 33204
Gijón, Spain (e-mail: [email protected]; [email protected];
load (Fig. 1), the variability of the conditions has a great impact
[email protected]; [email protected] ). on the main representative parameters of power quality.
Pablo Ardura G., Rocío Llera T. and Diego Cifrián R. are with the Global These electrical magnitudes are highly dependent of the
Research & Development Department. ArcelorMittal, Avilés, Asturias, 33400,
Spain. (e-mail: [email protected];
rolling process conditions, the chemical composition of the
[email protected]; [email protected] ). steel slab and the mechanical characteristics that the final coil

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 2

Fig. 2. Schematic of the rolling mill train

TABLE I
Fig. 1. Evolution of measured line current rms value during coil rolling MAIN CHARACTERISTICS OF THE STANDS
is intended to have. Hence the interest in developing an
MOTOR
TRANSFORMER
analytical method capable of reproducing the particular STANDS POWER POLES CYCLOCONVERTER
POWER (MVA)
(MW)
operating conditions of a hot rolling mill and predicting the
F1 8.0 6 Circulating-current-free 14.4
evolution of the main electrical variables that are involved.
F2 8.0 6 Circulating-current-free 14.4
B. Daily prediction of hot strip mill consumption
F3 8.0 6 Circulating-current-free 14.4
Daily prediction of large facilities consumption allows for
F4 8.0 6 Circulating-current-mode 15.7
significant savings in the cost of electricity. If the energy
market liberalization is taken into account, a prediction can be F5 8.0 6 Circulating-current-mode 15.7
leveraged to improve the negotiation with the utility and F6 8.0 6 Circulating-current-mode 15.7
obtain more competitive prices in the electricity market. Top Roughing 7.5 12 Circulating-current-free 14.4
Therefore, it is crucial to monitor and record (e.g. by using
Bottom Roughing 7.5 12 Circulating-current-free 14.4
a SCADA system) the electrical variables of the process along
with its key operating parameters to predict their influence on Circulating-current control allows for the increase of the
the electrical consumption. demand of reactive power when the plant operates at low load.
However, the high variability of the operating conditions of Since the 1970s, synchronous motors fed by thyristor-based
the process can hamper the drawing of conclusions regarding cycloconverters controlled by field oriented control (FOC)
the independent influence of each variable. Therefore, models have been extensively used in hot rolling mills. Since the
that reproduce the operation of the facility as closely as 1980s, the trends in steel mill drives have been to use pulse
possible are advisable. Such models should be compared to width modulation (PWM)-based voltage source inverters.
actual variables and validated afterwards. Once validated, the Despite the technological improvement that the use of
models become an important source of information to identify inverters involved, in the 1990s, up to ten hot rolling mill lines
trends in consumption. These trends can be analyzed by based on cycloconverters were installed. This technology is
varying a selection of influential variables; such a study is used today by many rolling mill plants [4], [5].
very complicated when conducted just on the basis of field The development of new high-power semiconductors and
measurements. multilevel inverter topologies has led to an increased
application of PWM-controlled voltage source inverters (VSI)
II. HOT ROLLING MILL ranging from 0.5 MVA to approximately 30 MVA. Converters
for steel mill drives must achieve good dynamic properties and
A. Roughing and finishing mills a low torque ripple, control reactive power consumption and
The modeled hot rolling mill is a classic facility mainly harmonic current injection, and be characterized by a high
comprising one roughing and six finishing mill stands (Fig. 2). efficiency and adjusted complexity and dimensions. In
The behavior of additional equipment such as coiling addition to the application of FOC in inverter-fed motor drives
machines, edgers, crop shear, etc. has not been reproduced in with various PWM schemes, such as carrier-based, hysteresis
the model because the major consumption of electricity is band control and space vector modulation, the recent
associated with the roughing and finishing mills. application of direct torque control (DTC) to AC drives in
The roughing mill consists of two stands, one at the top side plate rolling mills has been claimed as achieving the highest
and the other one at the bottom side. The drives are torque and speed performance ever attained with variable
circulating-current-free and double-cascade-connected speed drives, making it possible to control the full torque
cycloconverters. The double-cascade configuration allows for within a few milliseconds and reduce the impact of load
a better power factor by means of asymmetric and bias voltage shocks.
control. The rated data for the main equipment are shown in
Table I. B. Distribution network topology and filtering system
The finishing mill consists of six stands, three of them (F1, The hot rolling mill is connected to a 138 kV, 60 Hz,
F2 and F3) being driven by means of circulating-current-free, 3,600 MVA power distribution network through a 75 MVA,
double-cascade-connected, 12-pulse three-phase
138/34.5 kV transformer. The eight stands of the hot rolling
cycloconverters. The other stands (F4, F5 and F6) are driven
mill are fed from a 34.5 kV bus. Four banks of passive
from circulating-current-mode, 12-pulse three-phase
filtering are also connected to these same bars. The filtering
cycloconverters.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 3

system comprises one C-filter, two second-order high-pass

filters and one tuned filter. The global passive filter enables
the compensation for 45 Mvar at 34.5 kV (Fig.3). The
passive harmonic filter has been designed to improve the
harmonic response and provide the reactive power
compensation [6].
As shown in Fig. 4, Hcds(s) is the transfer function of a
current-divider that considers the relationship between the
power system current, Is, and the current injected from the
load, I,
I s ( s) Z f ( s)
H cds ( s)   . (1)
I ( s) Z f ( s)  Z s ( s) Fig. 3. Single-line diagram distribution network and filtering system
Fig. 5 shows the attenuation provided by the filtering
system in coordination with the impedance of the power
distribution system. The transfer function Hcds(s) is based
on the single-phase equivalent circuit assuming that the
system is balanced [7]. The attenuation factor of the filter
bank is mainly effective for high frequencies. Therefore,
although high-frequency components are attenuated, some
inter-harmonics on the low frequency side are still present Fig. 4. Current divider based on the filter/system impedance
and, in some cases, could be amplified due to parallel
resonances.

III. ANALYSIS METHOD

The developed analysis method is aimed to identify the
key variables influencing the electrical behavior of the main
stands of a hot rolling mill plant. This process is based on
both the characteristics of the steel to be milled and the Fig. 5. Frequency response of Hcds(s)
particular features of the rolling mill. The method provides
the evolution of the main electrical variables that are The FOC method is used to control the synchronous motor
involved. [10]. This strategy pursues the decoupling of flux and torque
control. Therefore, the flux can be mostly fixed by the field
The demand of active and reactive energy, the
windings while the armature current is used to achieve the
consumption of active and reactive power, and the current
motor torque. The motor torque is set through a speed control
harmonic injection into the power distribution system are the
loop that tracks the speed reference and takes the load torque
main variables of interest to be calculated. The results have
as a disturbance. All of the main loops of the FOC method
been compared to the corresponding measurements in an
have been modeled (Park transformations, flux observer,
actual rolling plant and then validated.
armature current, field current and speed loops, etc.). There
The main stands of the mill, the power network and the
are three different control loops:
filter banks have been modeled by using a program for
simulating electrical power systems [8]. Each stand has been - Speed loop: To set the flux and motor torque references
modeled by including a 33/1.15 kV step-down transformer, after tracking the speed reference and performing the field
which converts the 6-pulse topologies of the individual weakening.
rectifiers to twelve-pulse cycloconverter topologies. Each - Flux loop: To estimate the flux position through an
stand includes six transformers (two transformers per phase). observer and control the field current.
Cycloconverters with their corresponding control stage have - Current loop: To track the armature current components
also been modeled [9] (bias control, asymmetrical control, (mostly torque current).
circulating current control, etc.), as well as the synchronous
motor vector control that determines the modulation index and The cycloconverter firing signal generator has also been
the phase angle of the control signal. modeled. The voltage reference set by the motor control and
The load torque developed by the rollers of each stand is the currents are the inputs of the generator and are used for the
inhibition of the current-free cycloconverter firing pulses.
applied to the motor (Fig. 6). The control system uses the
speed as a reference (Fig. 7). Both variables are subject to the Furthermore, the generator uses the cosine-wave crossing
necessary adaptations to the speed-reducer gearboxes and the technique for the gate signal generation and other secondary
consequent performance ratio. Therefore, torque and speed are control techniques for enhancing the reactive power
the inputs to the developed electric model. consumption. Such control techniques are:

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 4

V. MILL PROCESS. INFLUENCING VARIABLES

The slab reaches the roughing mill free of mill scale, where
it is given the correct thickness to enter the finishing mill train.
Five to seven passes are run in the four-high reversing stand to
roll the slab down to the required transfer-bar thickness. The
number of passes depends on the measurements of the slab,
the steel plate to be obtained, the grade of steel and the final
rolling temperature in the train. The steel plate leaving the
Fig. 6. Calculated electromagnetic torque vs. measured load torque in stand roughing mill train is subject to a single pass in the continuous
F2
finishing train, then partially cooled but coiled while still hot.
The finishing mill train consists of six nonreversible two-high
arrangement stands where the required final thickness is
achieved. The temperature varies between 800 ºC and 1,250 ºC.
Hot rolling must be conducted at the appropriate temperature,
both on the surface and inside the material [14], [15].
For the determination of the mechanical conditions of the
rolling mill, mathematical models able to interrelate the
process variables with the characteristics of the steel sheet are
used. The models provide data about reductions in stands, roll
speeds, temperature at the end of the last stand, cooling rate,
Fig. 7. Speed control in stand F2; calculated motor speed vs. measured reference coiling temperature, etc. Several programs allow for
speed estimating the rolling conditions, such as StripCam and
- Asymmetric control is a strategy used in cascade HSMM (Hot Strip Mill Model). Both programs are capable of
cycloconverters based on the non-identical firing signals interrelating the rolling process variables with the mechanical
between the two series-connected bridges. The effect is an properties of the steel to be obtained [15], [16].
asymmetric firing in the cascade bridges. It has been
implemented in the circulating-free-current VI. SYNCHRONOUS MOTORS LOAD TORQUE AND SPEED
cycloconverters (R1, R2, F1, F2 and F3).
Rolling conditions for each type of steel are described as
- Reactive control through circulating current aims to follows. The required load torque and speed of the motors are
consume a higher reactive current in the circulating- calculated from summarized bellow conditions.
current-mode cycloconverters (F4, F5 and F6) when the
overall reactive consumption of the plant is low (e.g. in the A. Flat rolling
standstills) and the filters overcompensate for the reactive In flat rolling [17], the thickness is reduced by an amount called
power. the draft,
- Bias control aims to achieve voltages with higher rms
d  t0  t f , (2)
values at the converter output through third harmonic
injection. This kind of control leads to a higher voltage where d is the draft, t0 is the starting thickness and tf is the
level feeding the motor and, therefore, a decrease of the final thickness.When selecting the draft it is necessary to
input reactive power. optimize the mill operation within the limits set by the rated
power of the motors. The true strain, , undergone by the work
IV. TYPES OF STEEL. CHEMICAL AND ALLOY COMPOSITION in rolling depends on the previous and subsequent stock
thicknesses,
Three grades of steel with different carbon content are
considered, thus involving different hardness characteristics. t0 .
Steel types “A”, “B” and “C” have low, medium and high
  ln (3)
tf
carbon content, respectively. Characteristics of steel are
documented in AISI/SAE standard (American Iron and Steel The average flow stress, Y f , applied to the work material
Institute / Society of Automotive Engineers). The chemical in flat rolling, is determined from the true strain,
composition of each type of steel and its physical and
K  n
mechanical properties can be obtained from the Yf  . (4)
aforementioned standard (Tables II and III). Standard SAE 1 n
J403 1010 [11] is used for type “A” steel and Standard SAE Values of the strength coefficient, K, and the strain-
J403 1012 [12] for type “B” steel. The SAE standards between hardening exponent, n, depend on the composition, the heat
1005 and 1015 are used for steel whose carbon content is treatment and the work hardening. The higher the carbon
between 0.06 and 0.18%. Standard SAE J403 1050 [13] is used content is, the higher the strength coefficient results (Fig. 8).
for type “C” steel. SAE standards between 1035 and 1053 are The average flow stress is used to estimate the rolling force
used for steel whose carbon content is between 0.32 and 0.55%. and power.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 5

TABLE II
CHEMICAL COMPOSITION OF THE THREE TYPES OF STEEL
Element Steel A - Content (%) Steel B - Content (%) Steel C-Content (%)
Iron 99.3-99.7 99.16-99.6 98.46-98.92
Manganese < 0.6 0.3-0.6 0.6-0.9
Carbon < 0.12 0.1-0.15 0.47-0.55
Sulfur < 0.045 < 0.05 < 0.05
Phosphorous < 0.045 < 0.04 < 0.04

TABLE III
MECHANICAL PROPERTIES OF THE THREE TYPES OF STEEL
Properties Steel A Steel B Steel C
Tensile strength [MPa] 325 370 690
Yield strength [MPa] >280 310 580
Elastic modulus [GPa] 210 190-210 190-210
Poisson´s ratio 0.27-0.3 0.27-0.3 0.27-0.3
Elongation at break [%] 28 (in 80 mm) 19 (in 50 mm) 10 (in 50 mm)
Hardness, Vickers 105 108 207
Density [gr/cm3] 7.872 7.87 7.85

An approximation of the rolling force, F, can be calculated

from the average flow stress,
F  Yf  w L , (5) Fig. 8. Typical values of K and n for different types of hot rolled steel [18]

where w is the material width and L is the contact length, TABLE IV

which considering R the roll radius is approximately GEOMETRY OF THE SLAB FOR DIFFERENT TYPES OF STEEL

L  R  (t 0  t f ) . (6) Element Steel A Steel B Steel C

Slab length (m) 10 11 11
The power, P, required to drive two powered rolls is twice
Slab width (mm) 1,259 1,224 1,248
as much as the product of the torque for each roll and the
Slab weight (ton) 22.75 23.78 24.68
angular velocity. The torque for each roll, T, is
T  0.5  F  L (7)

and the active power is

P  F  L  r , (8)

where  r is the rotational speed in radians per second. The

active power demanded by the motors must be correctly
distributed by choosing the strip thickness at the delivery side
of each stand. Temperature and deformation resistance play an
important role in the distribution of forces in the stands.
B. Steel types: rolling conditions
The behavior of the three types of steel has been obtained
from their chemical composition (see Table II). Fig. 9 shows
K-strength coefficient and n-strain-hardening exponent at the
rolling temperature during the corresponding passes. The strip
thickness at the delivery side of each stand has been chosen
(Figs. 10 and 11) by considering the constraints on the power
and speed of the stands and the measurements of the slab and
the rolls (Table IV).
The average flow stress is obtained from (4) and, then, the Fig. 9. Top: K-strength coefficient. Bottom: n-strain hardening exponent
force and torque applied by each roller are calculated from (5)
and (7). Constraints to the speed/frequency and maximum
power at the stands must be considered to set the speed of the
rolls. The mechanical conditions in Figs. 12 and 13 correspond
to rolls data. Speed limits for stands F1 and F2 are observed in
Fig. 14. These stands are subject to higher torques in the
finishing mill and, therefore, the constraints on speed are more
severe. The tensile strength obtained from the estimation of the
mechanical rolling conditions is compared to that provided by
the standard (Table V). Fig. 10: Calculated thickness of the slab during the passes in roughing mill.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 6

Fig. 15. Measured load torque at the stands during steel milling.

Fig. 11. Calculated thickness of the slab during the passes in finishing mill.

Fig. 16. Measured motor speed in the stands of the rolling mill during steel milling.
TABLE V
VALIDATION OF THE TENSILE STRENGTH FROM SIMULATION

Tensile strength Steel A Steel B Steel C

Simulation [MPa] 355 352 661

Standard [MPa] 325 370 690

Fig. 12. Top: calculated roll speed during the rolling of steel types A, B and
C. Bottom: calculated rolling force during steel mill.
VII. POWER AND ENERGY DEMAND
The demanded active power is directly calculated from the
values of the torque and speed of the rolling mill motors, the
performance ratio of the drives and the transformer and stand
line losses. Reactive power values are mainly obtained from
the cycloconverter mean trigger angle (Fig. 17). All
represented records in this section have been calculated. The
analysis procedure has been validated in section VIII using on-
Fig. 13. Calculated load torque of the roughing mill stands during milling of
different types of steel.
site measurements. This angle depends on the load angle, 0,
and the cycloconverter modulation index, m. The relationship
among the load angle, the modulation index and the
displacement factor at the input, cos(Φi), is as follows [19]:
m . (9)
cos(i ) 
cos(0 )  0  sin(0 ) 2
(1  4  ( ) )
  cos(0 )
In the case of synchronous motors fed from the
cycloconverters, the output displacement factor is close to unity
because vector control forces the synchronous motor to operate
Fig. 14. Speed limits of stands F1 and F2 vs. selected and calculated speed. with unity power factor. The highest input displacement factor
is reached when the output voltage is maximum and, therefore,
Estimation is validated because the values of the tensile the modulation index is equal to 1 [20], [21]. At low-modulation
strength have the same order of magnitude. The evolution of the index operation (low load), higher values of reactive power are
load torque and the motor speed at the stands of the roughing absorbed by the cycloconverter, whereas the active power
and finishing mill drives, corresponding to motor conditions, are absorbed by the cycloconverter increases and the reactive power
shown in Figs. 15 and 16. In the case of the finishing mill the decreases with a high modulation index [22]. The reactive
ratios of the gearboxes incorporated into the stands must be power, Q, is estimated from the active power and the input
considered: F1=5.2; F2=4.5; F3=3.0; F4=1.478: F5=1.0; displacement factor,
F6=1.0. The roughing mill does not incorporate any
Q  P  tan(i ) . (10)
gearboxes.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 7

Fig. 18. Evolution of reactive power demands at stand F4.

Fig. 17. Flow diagram for calculating reactive power consumption.

The reactive power demand in stands F4, F5 and F6 is

controlled by the active control of reactive power [23], [24].
Overcompensation in the distribution network is avoided at
low-load intervals because the extra reactive power from the
filter banks is demanded by these stands. These stands are
forced to consume an extra reactive power, Q0, to compensate
for the reactive power injection from the filter banks. Q0 is Fig. 19. Active and reactive power demanded by top roughing mill drive
calculated by subtracting the reactive power injected by the
different filter banks and the reactive power consumed by the
stands, which is associated with the particular loading
conditions. Such a difference is divided by 3 when positive in
order to distribute it among the three stands with circulatory
current. Each stand will provide a maximum value of reactive
power depending on the maximum power that the converter
can drive and the active and reactive power associated with
the loading conditions.
Fig. 18 shows four kinds of reactive power: the reactive
power associated with the particular loading conditions, the
maximum reactive power that the drive can deliver through Fig. 20. Active power demanded by the finishing mill.
the circulating current, the reactive power consumption
associated with the circulating current and the final demand C. Global demand
of reactive power associated with stand F4. Possible overlapping of roughing and finishing mill trains
must be considered for a global analysis of the demand. When
A. Roughing mill overlapping does not occur, the consumption is sequential and
the demands of active and reactive power are those
In this section, a comparative analysis of the active and corresponding to these trains, although not synchronized.
reactive power consumption in the roughing mill has been
Fig. 21 shows the active and reactive power flow
performed (Fig. 19). Type-C steel requires 7 passes.
downstream and upstream of the point of common coupling
Steel types A and B require only 5 passes. The maximum PCC .These variables are used to estimate the energy demand
consumption is limited by the maximum power that the drives (Fig. 22) and the evolution of the rms value of the current and
can deliver. Both active and reactive energy increase with the voltage (Fig 23). Knowledge of the evolution of the rms value of
steel hardness. The demand of active energy for type-C steel is the voltage at the coupling point is crucial to ensure compliance
39% greater than that required for type-A steel. with the requirements set by standards [25].
B. Finishing mill The frequency components of the currents demanded by the
The study conducted for the roughing mill is also performed stands are dependent of the number of converter pulses and the
for the finishing mill train (Fig. 20). In this case, the maximum output frequency of the cycloconverter, and, the motor speed. The
values of active power demand have the same order of amplitudes and phase angles of these components are quantities
magnitude and the difference between the types of steel relies that depend fundamentally on the analyzed frequency, the
on the duration of the mill period. Fig. 19 shows a difference of 20 modulation index of the cycloconverter and the filtering
s over a maximum period of 120 s. inductive impedances of the circuits [21].

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 8

Fig. 21. Active and reactive power demanded by hot rolling mill.
Fig. 24. Evolution of the voltage THD at the PCC.

Fig. 22. Active and reactive energy demanded by hot rolling mill.

Fig. 23. Evolution of the rms value of the current upstream of the PCC and of
the phase voltage at the PCC.

The frequency components from the stands, along with the

frequency response of the filter stage and the distribution
network (Fig. 5), allow for performing the analysis of the
dynamic evolution of the total harmonic distortion of the
voltage at the coupling point (Fig 24).
Fig. 25. Simplified flowchart to estimate the main variables.
VIII. VALIDATION OF THE RESULTS
The validation of the electric model is demonstrated by
comparing its results with on-site measurements during the
rolling of type “A” steel. To obtain the calculated variables, the
rolling process is simulated and the main variables are estimated
according to the flowchart shown in Fig.25 [26].The evolution of
the main electrical variables for the same type of steel is shown in
Figs. 26-28, which is obtained from on-site measurements, is
compared with the similar evolution of the calculated variables.
The sampling frequency is 25 kHz for the calculated variables
and 100 Hz for the measured variables. The measured THD of the Fig. 26. Validation of active and reactive power demanded by rolling mill.
voltage is a time-averaged record, whereas the calculated THD The achieved adjustment as shown in the Figs. 26-28 is
has been obtained at a higher registering frequency. Because accurate. The major discrepancies occur at periods of low load,
torque and speed, which feed the model, experience a significant due to the fact that only the most significant loads have been
change in the process that is reproduced, the electrical variables modeled (finishing and roughing mill) and the impact of these
also experience significant dynamic variations. discrepancies is not relevant to the overall analysis.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 9

parameters. A simple algorithm has been developed to set a

correlation between the mill operation conditions and the
electrical demand of the facility. The influence of several
mechanical variables and electrical variables on power
demand and power quality has been analyzed.
REFERENCES
[1] E. Kun, T. Szemmelveisz, “Energy efficiency enhancement in the Hot
Rolling Mill” Materials Science and Engineering, vol. 3, no. 2, pp. 43-
50, 2014.
[2] University of Oviedo, Arcelormittal Spain, Centre de Recherches
Metallurgiques ASBL, Vdeh-Betriebsforschungsinstitut Gmbh,
“Minimissing energy loss in hot rolling by intelligent manufacturing.”
European Commission, Research Fund of Coal and Steel. Tech Rep.
UE-10-RFSR-CT-2010-00008, Dec 2013.
[3] Chen, C. S., Y. D. Lee, C. T. Hsu, D. S. Ting, C. C. Shen, “Power
Quality Assessment of a Hot Strip Mill with Cycloconverter Drive
Systems,” Industry Applications Conference 2007, 42nd IAS Annual
Meeting, pp. 9-16, Sept. 2007, New Orleans LA.
[4] H. Hosoda, S. Kodama, R. Tessendorf “Large PWM Inverters for
Rolling Mills” Association for Iron & Steel Technology Digital Library
January 2008.
[5] A. K. Chattopadhyay, “Alternating Current Drives in the Steel Industry,
Advancements in the last 30 years,” Industrial Electronics Magazine,
vol 4, pp. 30-42, December 2010.
Fig. 27. Validation of the rms value of the current upstream of the PCC and [6] A. B. Nassif, W. Xu, “Passive Harmonic Filters for Medium-Voltage
of the phase voltage at the PCC. Industrial Systems: Practical Considerations and Topology Analysis,”
39th North American Power Symposium NAPS 2007, Sept. 2007 pp. 301-
307, Las Cruces NM.
[7] J. K. Phipps “A Transfer Function Approach to Harmonic Filter Design”
Industry Applications Magazine, vol. 38, no. 2, pp. 68-82, March/April
1997.
[8] MathWorks, 2010, The MathWorks—MATLAB and Simulink for
Technical Computing [Online]. Available: http://www.mathworks.com
[9] Wolfgang Timpe, “Cycloconverter Drives for Rolling Mills” IEEE
Trans. on Industry Applications, vol. IA-18, no. 4, pp. 400-404,
July/August 1982.
[10] J. R. Rodríguez, J. Pontt, P. Newman, R. Musalem, H. Miranda, L.
Morán, G. Alzamora, “Technical Evaluation and Practical Experience of
High-Power Grinding Mill Drives in Mining Applications” IEEE Trans.
on Industry Applications, vol. 41, no. 3, pp. 866-873, May/June 2005.
[11] AISI/SAE Standard (American Iron and Steel Institute / Society of
Fig. 28. Validation of the voltage THD at the PCC.
Automotive Engineers). SAE J403 1010. June 2014.
[12] AISI/SAE Standard (American Iron and Steel Institute / Society of
IX. CONCLUSION Automotive Engineers). SAE J403 1012. June 2014.
The relationship among the properties of the steel to be [13] AISI/SAE Standard (American Iron and Steel Institute / Society of
milled, the mechanical conditions of the rolling mill, the Automotive Engineers). SAE J403 1050. June 2014.
[14] J. L. E. Berciano, E.T. Guerra, S. E. de Bengy, D. F. Segovia
technical features of the roughing and finishing mills and the “Laminación. Monografías sobre Tecnología del Acero. Parte IV”
involved electrical variables has been obtained. The proposed Madrid, pp 36-49, January 2010.
method provides a powerful analytical tool for the prediction [15] F. Yamada, K. Sekiguchi, M. Tsugeno, Y. Anbe, Y. Andoh, C. Forse,
M. Guernier, T. Coleman “Hot Strip Mill Mathematical models and Set-
of the electrical demand and the power quality parameters. Up Calculation,”. IEEE Trans. on Industry Applications, vol. 27, no. 1,
The continuous variability of the process operating conditions pp. 131-139, January/February 1991.
can hinder the drawing of conclusions from actual [16] U. Lotter, H.-P. Schmitz, L. Zhang, “Application of the Metallurgically
Oriented Simulation System “TKS-StripCam” to predict the properties
measurements regarding the independent influence of each of Hot Strip Steels from the Rolling Conditions,” Advanced Engineering
variable. Materials,” vol. 4, Issue 4, pages 207–213, April, 2002.
A complete hot strip mill facility has been modeled. Each [17] Mikell P. Groover “Fundamentals of Modern Manufacturing. Materials,
stand of the roughing and finishing mills has been simulated. Processes and Systems” John Wiley & Sons, 4th edition, pp. 395-403,
2010.
The electric distribution system of the factory with the existing [18] “ASM Metals Handbook, vol. 14, Forming and Forging,” ASM
passive filtering stage has been modeled as well. The model International, pp 351-355, 1996.
has been tested with several load profiles by using the torque [19] J. M. D. Murphy, F. G. Turnbull “Power Electronic Control of AC
Motors” Pergamon Press, pp 199-216, 1989.
and speed variables as input data, and with on-site [20] P. Aravena, L.Morán, J. Dixon, J. Espinoza, O. Godoy “A New Hybrid
measurements of voltage, current, active and reactive power, Filter Topology for Sub and Inter-harmonic Attenuation in
torque and rolling speed. A database with three steel types has Cycloconverter-Fed Drives Applications” Industry Applications Society
Annual Meeting 2009, IAS 200, 4-8 Oct 2009, Houston, Texas.
been analyzed, which allows for correlating the type of steel [21] B. R. Pelly, Thyristor Phase-Controlled Converters and
with the electrical demand and several power quality Cycloconverters. Hoboken, NJ, USA: Wiley-Interscience, 1971.

0093-9994 (c) 2015 IEEE. Personal use is 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/TIA.2016.2533483, IEEE
Transactions on Industry Applications
>ID 2015-METC-0373.R1 < 10

[22] P. Aravena, L. Morán, R. Burgos, P. Astudillo, C. Olivares, and D. A. José M. Cano (M’98) was born in
Melo “High-Power Cycloconverter for Mining Applications: Practical
Oviedo, Spain, in 1971. He received the
Recommendations for Operation, Protection, and Compensation” IEEE
Trans. on Industry Applications, vol. 51, no. 1, pp 82-91, M.Sc. and Ph.D. degrees in electrical
January/February 2015. engineering from the University of
[23] S. G. Bosga, J. L. Duarte, A. J. Vandenput, “Natural Circulating Current Oviedo, Spain, in 1996 and 2000,
Control of a Cycloconverter,” Industry Applications Society Annual
respectively.
Meeting, 1993.Conference Record of the 1993 IEEE, pp. 1160 – 1165,
vol 2, October 1993 Toronto, Ont.. In 1996 he joined the Department of
[24] S. Tanaka, “Control Method for Cycloconverter and control Apparatus Electrical Engineering, University of
Therefor,” U. S. Patent 4,670,826. Jun, 2, 1987. Oviedo, where he is currently an Associate Professor. During
[25] “IEEE Recommended Practices and Requirements for Harmonic Control
in Electrical Power Systems” IEEE Std 519-2014, 2014.
2012 he held an appointment as a Visiting Associate Professor
[26] G. A. Orcajo, J. Rodríguez D., P. Ardura G., J. M. Cano, J. G. at the Department of Electrical and Computing Engineering,
Norniella, R. Llera T., D. Cifrián R. “Dynamic Estimation of Electrical University of British Columbia, Canada. His main research
Demand in Hot Rolling Mills” IEEE Industry Applications Society interests are in the field of power quality solutions for
Annual Meeting 2015, 18-22 Oct 2015 ,Dallas, Texas.
industry, power converters, distributed generation and smart
Gonzalo A. Orcajo was born in Gijón, grids.
Spain in 1965. He received the M. Sc. and
Ph. D. degrees in electrical engineering Joaquín G. Norniella was born in Gijón,
from the University of Oviedo, Spain, in Spain, in 1980. He received the M.Sc. and
1990 and 1998, respectively. the Ph.D. degrees in electrical engineering
In 1992, he joined the Department of from the University of Oviedo, Spain, in
Electrical Engineering, University of 2005 and 2012, respectively.
Oviedo, where he is currently an He worked as a research and development
Associate Professor. His main research interests are in the engineer at the Department of Electronic
field of power quality in industrial power systems. In recent Engineering of the University of Oviedo
years, he has focused on the detection and location of faults in for twenty months since February, 2007.
distribution systems and on reactive power compensation In 2008 he joined the Department of Electrical Engineering of
systems. the aforementioned University, where he is currently working
as a Lecturer. His main research interests are in the field of
Josué Rodríguez D. was born in Pravia, power quality solutions for industry and power converters.
Spain, in 1989. He received the M. Sc. in
electrical engineering from the University Rocio Llera was born in Oviedo, Spain,
in 1983. She received the M.Sc degree in
of Oviedo, Spain in 2014.
Forest & Environment Engineering from
He worked as a research fellow at the
the University of Oviedo, Spain, in 2006.
Department of Electrical Engineering of
She is currently pursuing the Ph.D degree
the University of Oviedo between
in Mining, Civil Engineering and Project
October 2013 and June 2015. In July
Management.
2015, he joined the Department of Beams-RadioFrequency-
Power amplifiers and Modules production (BE/RF/PM) at She worked as a research engineer at the
European Organization for Nuclear Research (CERN), Mining Exploitation Department of the
Geneva, Switzerland. He works on the construction and University of Oviedo, for four years since 2006, on
commission of electronic test benches for future medical environmental and energy topics in the steel sector. In 2010
applications. His main research interests are in the field of she joined ArcelorMittal, where she is currently working as
reactive power compensation systems and power quality. R&D engineer in the Global R&D Center in Asturias. She is
involved in energy issues focused on process modeling, waste
Pablo Ardura was born in Luarca, Spain, energy recovery and energy efficiency.
in 1986. He received M.Sc. degree in
electrical and electronic engineering from Diego Cifrián R. was born in Oviedo,
the University of Oviedo, Spain, in 2010. Spain, in 1987. He received the M.Sc
He is currently pursuing the Ph.D degree degree in electrical engineering from the
in electrical engineering. From 2010 to University of Oviedo, Spain, in 2010.
2014 he worked at the Department of In 2014 he joined ArcelorMittal, where
Electrical Engineering of the University he is currently working as an R&D
of Oviedo. In 2015 he joined the ArcelorMittal Global R&D engineer in the Global R&D Center in
Center in Asturias, where he is currently working as a research Asturias. His main research interest is in
engineer. His main research interests are in the field of power the field of innovative solutions for the steelmaking industry,
quality, power converters, energy storage and waste-heat focused on electrical and combustion process improvement.
recovery systems for the steelmaking industry.

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.