A TORQUE ESTIMATION METHOD TO AID AN INTELLIGENT

MANAGEMENT SYSTEM FOR OIL WELLS AUTOMATED

Alberto S. Rebouc¸as, Fl

avia N. Seraﬁm, Milena de A. Moreira, Ven

ıcio R. V. Rodeiro,

Amauri Oliveira and J

es J. F. Cerqueira

Departamento de Engenharia El

etrica of the Escola Polit

ecnica

of the Universidade Federal da Bahia

Rua Aristides Novis, 02, Federac¸

ao, CEP:40210-630 Salvador, Bahia, Brasil

Keywords:

Intelligence, Supervision, Estimation, Induction Motors.

Abstract:

This article presents a contribution for an oil wells intelligent management system called SGPA that nowadays

manages about 700 oil wells using the rod pumping lift method at Bahia State, Brazil. The intelligent man-

agement system will be applied on oil wells using the gradual pumping method. In this oil pumping method,

the torque on the rod is very important for detection of operational problems. It will be considered that the

well is driven by an induction motor. A torque estimation method on rod and some results from laboratory are

presented.

1 INTRODUCTION

In oil reservoirs, usually the pressure of bottom is not

enough to move oil to surface, either for its natural

characteristic or the end of its productive life. In this

case, artiﬁcial lift systems are used to move oil to sur-

face. To select the ﬁttest lift method, it is necessary to

know the conditions of the wells and to take in con-

siderate several factors:

• Characteristics of the well such as porosity, perme-

ability, sand rate, gas rate, pressure, temperature,

production forecast, diameter of the covering and

depth;

• Recovery methods for the well due to alterations of

behavior during the operational life;

• Fluid properties such as density, viscosity, parafﬁn

rate and sand rate;

• Energy available such as electric energy, gas or

fuel;

• Localization, number of wells and resources from

the ﬁeld;

• Legal restrictions such as ambient standards;

• Economic evaluation of the project;

The main artiﬁcial lift methods are:

Continuous and Intermittent Gas-lift: The gas-lift

(CGL or IGL) uses the energy of gas compressed

to move the oil to surface. The gas is used to gasify

the oil (GLC) or to throw it (GLI) of the deep until

the surface.

Submerged Centrifugal Pump: The Submerged

Centrifugal Pump consists of a centrifugal pump

of several stages in serie, in accordance with

the pressure of the bottom of the well. Motors

projected to work under high temperature and high

pressure are used into this lift system;

Rod Pumping: In the Rod Pumping (RP), an elec-

tric motor or a combustion motor drives an appara-

tus to convert the rotational movement into linear

movement to a rod. This rod drives a piston pump

at the bottom of the well.

Gradual Pump: The Gradual Pump (GP) uses a

pump with a steel screw rotor and a stator of soft

material, usually an elastomer. The pump is used

in the bottom of the well working submerged into

oil. The rotational movement of the pump dislo-

cates the oil gradually in the direction of the sur-

face. Without necessity of valves, the pump can be

driven from the surface with the rotational move-

ment transmitted by a rod. Electric motors can be

used to drive the pump.

An oil extraction ﬁeld can have a lot of wells. For

example, the Bahia Extraction Field, at Brazil, has

about 4000 wells. Such numbers of well produces a

lot of information such that its analysis by engineers

is very difﬁcult. Some times, this analysis are nec-

essary to intervene in the well operation. Specialists

270

Rebouças A., Seraﬁm F., Moreira M., Rodeiro V., Oliveira A. and Cerqueira J. (2004).

A TORQUE ESTIMATION METHOD TO AID AN INTELLIGENT MANAGEMENT SYSTEM FOR OIL WELLS AUTOMATED.

In Proceedings of the First International Conference on Informatics in Control, Automation and Robotics, pages 270-275

DOI: 10.5220/0001137802700275

 SciTePress

motor

head

rod

pump

Figure 1: Gradual pump system (GP).

have proposed the use of intelligent system approach

to aid the management of oil wells (Alegre et al.,

1993; Hosn et al., 2001; Nikravesh and Aminzadeh,

2001; Cerqueira et al., 2002b; Cerqueira et al., 2002c;

Cerqueira et al., 2002a; Cerqueira et al., 2003).

The torque variable is very important to supervise

the wells. The torque analysis supplies information

about the load of the electric motor. This makes pos-

sible preventive actions to avoid that the system oper-

ates in a dangerous situation.

The goal of this paper is to present a torque esti-

mation method to aid the supervision of GP in oil

well. The torque estimated at head of the well can

warn operational problems. The GP system is driven

by induction motor in the surface that transmits the

rotating movement for the pump at the bottom of the

well by a rod coupled with a gear train (see ﬁgure 1).

Currently, a frequency inverter can be used to control

the motor speed.

The paper is organized this way. In section 2 are

presented operational problems in GP. In section 3

is presented an intelligent distributed system used to

manage oil wells, the SGPA. In section 4 the torque

estimation method for induction motor is described

in mathematical format. The hardware/software ar-

chitecture used to estimate the torque is presented in

section 5. In sections 6 and 7 are presented some ex-

periment results and the conclusions of the paper re-

spectively.

2 OPERATIONAL PROBLEMS

Excess of torque on the rod that moves the pump in

the bottom of the well can break the rod or the gear

train coupled with motor and to stop the oil pumping.

For repair a broken pump/rod set is necessary to re-

move a rod with hundreds of meters and to capture

the pump fallen at bottom of the well using special

equipment and specialized people. Sometimes, the

well recovery can delay weeks and to cause losses of

production and of cash.

On the other hand, a very small torque on the rod

causes low power factor of power on the electric sys-

tem and can indicate that the pump is not pumping

the ﬂuid. This situation could damage the elastomer

of the stator hindering the perfect rotation of the rod

and leading the system to the previous situation, with

high torque. Additionally, when the ﬂuid is pumped,

it lubricates and refrigeration of the mechanical sub-

system of bottom of the well.

A previous fault detention allows the planning of

the maintenance of the lift system, in order to attenu-

ate the impacts provoked for the operation stops and

to reduce the maintenance costs.

It is necessary accompaniment of other variable of

the system such as the noise level in the head, the

pressure in the top of the well and the electric current

on the motor during the pumping of the oil. These

variable are also important for fault detection in lift

system, but they are not approached in this paper.

3 THE INTELLIGENT

MANAGEMENT SYSTEM

Researchers of Universidade Federal da Bahia in

Brazil joint with engineers of Brazilian petroleum

company (PETROBRAS) have applied concept of ex-

pert systems into an automated lift well management

system, called SGPA - Sistema de Gerenciamento

de Poc¸os Automatizados (Cerqueira et al., 2002b;

Cerqueira et al., 2002c; Cerqueira et al., 2002a;

Cerqueira et al., 2003). The SGPA consists of three

levels of management:

Level 1: Where a programmable logical controller

(PLC), sensors and actuators work directly in the

supervision and control of the well. The electric

motor is supplied by electronic inverter. This level

of management and automation is local and the

well become automated;

Level 2: Where intelligent supervisor system re-

ceives the information from level 1 by radio com-

munication link. The intelligent supervisor man-

ages group of wells of an extraction ﬁeld from a

control room;

Level 3: Where are made analysis, diagnostics and

recovery actions are chosen for the lift systems.

In this level is added information about production

costs, economic information, operational historical

of the wells, and information from previous levels.

In this management model, Artiﬁcial Intelligence

techniques such as Symbolic Neural Nets and Fuzzy

Logical are used for application of knowledge get

from specialists in petroleum engineering.

A TORQUE ESTIMATION METHOD TO AID AN INTELLIGENT MANAGEMENT SYSTEM FOR OIL WELLS

AUTOMATED

271

torque

rotation

Figure 2: Curve torque × rotation for induction motor.

For torque measurement from rod in GP method,

a torquimeter could be installed in the rod. However,

a torquimeter is not a robust equipment and has high

cost, mainly considering that must be installed in the

well surface and submitted to weather action. The

same inconvenience is observed when is analyzed the

possibility of insertion of a torquimeter in the gear

train or in the electrical motor or in the bottom of

the well. The insertion of a torquimeter in the mo-

tor would had advantage the torquimeter works with

lesser torques.

A torque estimation method developed for the in-

duction motor from the spectrum analysis of the elec-

tric current is presented as option for torque estima-

tion in the rod and to supervise the GP method. This

estimation method does not modify the mechanical

structure of the wells. Therefore, the method is not

intrusive. Petroleum engineers can use the torque es-

timated to study problemas in the wells.

The torque estimation from the current spectrum

offers satisfactory accuracy, practically without intru-

sion in the system, but it requires sophisticated anal-

ysis. The identiﬁcation of a speciﬁc frequency har-

monic component of the electrical current on the elec-

tric motor is necessary. This frequencies are function

of the rotor speed. With this, the slip of the induction

motor can be obtained. The slip has an almost linear

relation with the torque into operational region. From

this linear relation, the torque can be estimated.

4 THE TORQUE ESTIMATION

METHOD

In curve torque × rotation for induction motor is ver-

iﬁed that near to the synchronous speed the curve can

be approached by a straight line with negative angu-

lar coefﬁcient, as shown in ﬁgure 2, where: n

is the

nominal speed; τ

is the nominal torque; n

is the

synchronous speed; n is the speed; and τ is the torque.

From nameplate data, the torque can be estimated

by the expression

τ =

− n)

− n

)

0.10466 n

(1)

where P

is nominal power and the speeds are in rpm

(rotation per minute).

This expression is gotten from a other expression

used for estimation of the power output of the induc-

tion motor and indicated by technical standards (IEEE

Standard, 1991; Damasceno, 2002; Damasceno et al.,

2002a; Damasceno et al., 2002b). For torque esti-

mation, the information about electric motor speed is

necessary.

4.1 SPEED ESTIMATION

The speed measurement in induction motor is as dif-

ﬁcult as torque measurement in the GP system. Me-

chanical considerations and noise make the use of

tachometer or encoders connected to the axle of the

motor a solution does not appropriate. An alterna-

tive is the speed estimation from spectrum of the elec-

tric current signal of any electrical phases that supply

the motor, as described for (Hurst and Habetler, 1996;

Blasco-Gimenez et al., 1996; Benbouzid, 2000).

Because interactions between magnetic ﬁelds into

the induction motor and constructive imperfections -

as the slots in the rotor where the squirrel cage is built

- the spectrum of the electrical current is rich in har-

monic components. Some of these components are

related to the slip in the form

= f

⎡

⎢

⎣

(kR± n

)

⎛

⎜

⎝

1 − s





⎞

⎟

⎠

± n

⎤

⎥

⎦

(2)

where f

is electrical fundamental frequency, f

frequency multiple of the fundamental frequency, R

is the number of rotor slots, n

it is the eccentricity,

P it is the pole number of the stator, n

is a harmonic

order and k is a any multiple.

The method works as follows. For s =0, values for

k, n

and n

are attributed. With this, a frequency

called f

is determined. A maximum slip value

max

is attributed and applied in equation 2. This de-

termines a frequency f

min

. So, a spectral window

from f

min

to f

can speciﬁed where f

will be

always the maximum amplitude component. The mo-

tor slip can be determined from expression

s =1−



− n



kR± n

. (3)

The motor speed can be determined in rpm from

expression

n =(1− s)



120 f



. (4)

ICINCO 2004 - INTELLIGENT CONTROL SYSTEMS AND OPTIMIZATION

272

Current

Sensor

Current

Sensor

Data

Acquisition

Detector

of f

Filter 1

Detector

of f

Equations

1e2

Figure 3: System Blocks Diagram.

This estimation method has accuracy enough to be

used in torque estimation. Details about implementa-

tion of speed estimation method can be found in (San-

tos, 2002; Santos and Oliveira, 2002).

5 THE SYSTEM ARCHITECTURE

The systems built from three subsystem: one of data

acquisition; one of hardware for data processing; and

one third of software.

The data acquisition subsystem is composed of cur-

rent sensors and a data acquisition hardware with

sample-hold and analog-to-digital converter.

The processing subsystem is constituted of a per-

sonal computer. The software subsystem was imple-

mented using resources of the LABVIEW



package.

Figure 3 illustrates the organization of the system.

Two current sensors are used. A sensors is used for

current signal acquisition and subsequent detention

of electrical signal fundamental frequency, f

. The

other sensor is used for current signal acquisition and

subsequent detention of the frequency that carries the

information about the induction motor slip, f

For detention of f

, the current signal is ﬁltered

by an pass-band analog ﬁlter, represented in the ﬁgure

for ﬁlter 1. This ﬁlter objectives to eliminate in order

the fundamental component of the electrical signal,

and the high frequencies. In sequence, the data ac-

quisition is made. The digital signal is processed for

detection of the component f

. After the detention

of f

and of f

, the information about R, P

, n

and P

are used to estimate the speed and the torque

in the induction motor attributing values for k, n

and

into expressions (1)-(4).

From the speed reduction on the gear train the

torque in well head is gotten. Petroleum engineers

use this torque to estimate the torque at some points

of the rod. The SGPA can be implemented adding in

order these information with other about the working

of the wells.

6 EXPERIMENTAL RESULTS

Laboratory tests was done for validation of the the-

oretical developments presented in this text. In the

test system, a continuous current generator simulates

a mechanical load in the axle of the three-phase in-

duction motor corresponding to mechanical load on

rod in the GP system. For the torque estimation sys-

tem illustrated in ﬁgure 3 and described in section 5,

a tachometer and a torquimeter to compare its indica-

tions with estimated values were added.

The torque estimation system was tested with an

induction motor. A 2 cv, 220/360 V (∆/Y), 4 pole,

1725 rpm, and 44 rotor slots. For this motor, the ex-

perimental results are presented in table 1.

The ﬁrst measure was carried out with the contin-

uous current generator working without ﬁeld current

and with this the induction motor worked with very

small load. The last measure was carried out at over-

load condition. Analyzing tables 1 is observed that

inside normal band operation of induction motor the

error in the torque estimation is very small. The errors

were only signiﬁcant in order on overload and very

small load condition. Additionally, at overload con-

dition, the error can be attributed to linear approach

made for torque, once the true curve is not a straight

line. Therefore, the torque estimation method is efﬁ-

cient inside motors operation band.

The efﬁciency of the estimation method depends on

accuracy of data from manufacturer of the induction

motors. The nameplate data have their accuracy stan-

dardized, but this does not avoid that any induction

motor works outside the nameplate data. Some times,

nameplate data and manufactures book data are differ-

ent. So, part of error for the motor can be attributed

A TORQUE ESTIMATION METHOD TO AID AN INTELLIGENT MANAGEMENT SYSTEM FOR OIL WELLS

AUTOMATED

273

Table 1: Experimental Results for 2 cv Motor.

n measure (rpm) n estimated (rpm) τ measured(Nm) τ estimated (Nm) error (%)

59.992 1792 1790 0.82 1.01 23

60.005 1782 1782 2.02 1.94 3.96

59.978 1764 1763 4.03 3.95 1.99

59.978 1744 1744 6.03 6.04 0.17

60.031 1724 1723 8.08 8.28 2.48

60.005 1696 1696 10.05 11.23 11.74

60.024 1667 1666 12.01 14.44 20.23

to accuracy of manufacturer data.

7 CONCLUSION

The estimation methods of torque and speed are suf-

ﬁciently accurate inside operation nominal band of

three-phase induction motors. The use of these meth-

ods makes possible the GP supervision without the

installation of encoders, tachometers or torquimeters.

This means cost reduction of systems like SGPA in

GP wells.

A next step is the test of the system using frequency

inversor and its evaluation on the wells. It is intended

also improvement of hardware using a dedicated sub-

system. This subsystem must be built with microcon-

troller and digital signal processor (DSP). The com-

plete system will be used in the supervision of ap-

proximately 200 oil wells installed in Bahia State, in

Brazil.

ACKNOWLEDGEMENTS

The authors would like to thank in order the Finan-

ciadora de Estudos e Projetos (FINEP), the PETRO-

BRAS S. A., and the Conselho Nacional de Desen-

volvimento Cient

ıﬁco e Tecnol

ogico, all of Brazil, by

their support for this work.

REFERENCES

Alegre, L., Morooka, C. K., and Rocha, A. F. (1993). In-

telligent diagnosis of rod pumping problems. In Pro-

ceedings of the Annual Technical Conference and Ex-

hibition of the Society of Petroleum Engineers, pages

97–107, Houston, USA.

Benbouzid, M. E. (2000). A review of induction motors sig-

nature analysis as a medium for faults detection. IEEE

Transations on Industrial Electronics, 47(5):984–993.

Blasco-Gimenez, R., Asher, G. M., Summer, M., and

Bradley, K. J. (1996). Performance of FFT-rotor slot

harmonic speed detector for sensorless induction mo-

tor drives. IEE-Proceedings on Electronic Power Ap-

plied, 11(1):66–73.

Cerqueira, J. J. F., Corr

ea, J. F., Lepkison, H., Bitencourt,

A. C., and Schnitman, L. (2002a). Development of an

intelligent distribuited manegement system for auto-

mated wells. In Proceedings of the Annual Technical

Conference and Exhibition of the Society of Petroleum

Engineers, pages 3053–3058, San Antonio, USA.

Cerqueira, J. J. F., Jesus, A., Dantas, E., Pereira, C. H.,

Albuquerque, G., Bitencourt, A. C., Pereira, F., Lep-

kison, H., and Corr

ea, J. F. (2002b). An intelligent

distributed system applied on automated lift system.

In Proceedings of the WSEAS International Confer-

ence on System Science 2002, pages 1951–1955, Rio

de Janeiro, Brazil.

Cerqueira, J. J. F., Jesus, A., Dantas, E., Pereira, C. H.,

Albuquerque, G., Bitencourt, A. C., Pereira, F., Lep-

kison, H., and Corr

ea, J. F. (2003). An intelligent

distributed system applied on automated lift system.

WSEAS Transactions on Systems, 2(2):482–486.

Cerqueira, J. J. F., Lepkison, H., Corr

ea, J. F., Biten-

court, A. C., Reis, A. C., Jesus, A., Fisch, M., Dan-

tas, E., Pereira, C. H., Albuquerque, G., and Pereira,

F. (2002c). Sistema de gerenciamento de poos de

petr

oleo automatizados. In Anais do XIV Congresso

Brasileiro de Autom

atica, pages 3053–3058, Natal-

RN, Brazil.

Damasceno, S. M. N. (2002). Uma contribuio a avaliac¸

ao de

rendimento de motores de induc¸

ao trif

asicos. Master’s

thesis, Escola Polit

ecnica da Universidade Federal da

Bahia, Salvador-BA, Brazil.

Damasceno, S. M. N., Cerqueira, J. J. F., Lima, A. C. C.,

Oliveira, A., and Jesus, R. J. (2002a). Uma pro-

posta para avaliac¸

ao de rendimento de motores de

induc¸

ao em cho de fbrica. In Anais do XIV Congresso

Brasileiro de Autom

atica, pages 1136–1141, Natal-

RN, Brazil.

Damasceno, S. M. N., Oliveira, A., Cerqueira, J. J. F., Lima,

A. C. C., Santos, E. T. F., Cunha, V. J. G., and Jesus,

R. J. (2002b). Studies about yhe methodology for the

evaluation of energetic efﬁciency in trifasic induction

motors on ﬁeld. In Anais do V INDUSCON - Con-

ICINCO 2004 - INTELLIGENT CONTROL SYSTEMS AND OPTIMIZATION

274

fer

encia de Aplicac¸

oes Industriais do IEEE Brasil,

pages 582–586, Salvador-BA, Brazil.

Hosn, N. A., Popa, A. S., and Popa, C. G. (2001). A new re-

alistic approach to pumping unit optimization through

the use of intelligent system. In Proceedings of the

Eastern Regional Meeting of the Society of Petroleum

Engineers, pages 1–9, Canton, USA.

Hurst, J. S. and Habetler, T. G. (1996). Sensorless speed

measurement using corrent harmonic spectral estima-

tion in induction machine drives. IEEE Transations

on Power Electronics, 11(1):66–73.

IEEE Standard (1991). IEEE standard test procedure for

polyphase induction motors and generators 112-1991.

Nikravesh, M. and Aminzadeh, F. (2001). Mining and fu-

sion of petroleum data with fuzzy logic and neural net-

work agents. Journal of Petroleum Science and Engi-

neering, 29(2):221–238.

Santos, E. T. F. (2002). An

alise e desenvolvimento de

um ampliﬁcador lock-in digital. Master’s thesis, Es-

cola Polit

ecnica da Universidade Federal da Bahia,

Salvador-BA, Brazil.

Santos, E. T. F. and Oliveira, A. (2002). Estimac¸

ao indireta

de velocidade de um motor de induc¸

ao trif

asico uti-

lizando um analisador de espectro lock-in. In Anais do

XIV Congresso Brasileiro de Autom

atica, pages 304 –

308, Natal-RN, Brazil.

A TORQUE ESTIMATION METHOD TO AID AN INTELLIGENT MANAGEMENT SYSTEM FOR OIL WELLS

AUTOMATED

275