SPE 53962

Surface Axial Load Based Progressive Cavity Pump Optimization System

L Mena, SPE, and S. Klein, SPE, InterRep

Copyright 1999, Society of Petroleum Engineers Inc.

time and equipment useful life, while optimizing the well
This paper was prepared for presentation at the 1999 SPE Latin American and Caribbean production.
Petroleum Engineering Conference held in Caracas, Venezuela, 21–23 April 1999.

This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
Introduction
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to The Progressive Cavity Pump Optimization System
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at (PCPOS) is a set of equipment and computerized
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
mathematical model that determines important downhole
for commercial purposes without the written consent of the Society of Petroleum Engineers is values and then takes control actions on the well RPM and
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous diluent or chemical Injection apparatus to optimize production
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. and adapt the well to possible changing operating conditions.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
It is based on the surface axial load (as well as other surface
variables) as a primary value to be used in diagnostics of a
Abstract PCP well, because as shown here, axial load does have a direct
Actual Field tests performed in heavy and extra- relationship to downhole pump dynamics.
heavy oil wells in San Tome, Venezuela reveal that axial load PCPOS uses artificial intelligence software to control the
does have a direct relation with Progressive Cavity Pump well, an expert system that determines an operational pattern
(PCP) downhole conditions. The differential pressure across for the PCP. It also evaluates a set of rules based on input data
the pump is easy to infer by just measuring the surface axial coming from a well head instrumentation package that
load of the system and surface pressures. This fact is includes axial load supported by the bearing in the surface
particularly useful to diagnose a PCP (known to be very drive head, surface tubing pressure, surface casing pressure,
difficult to do for this system). Good inferences on downhole motor current and torque.
conditions allow PCP optimization to be done much easier and Tests of PCPOS initially started in 1996 in well MEL-39
more efficiently. It has been proven in beam pump in San Tome, Eastern Venezuela. These tests were continued
applications that the dynagraph chart (axial load vs. pump in several other wells. The expert system software program
cycle) can be used to diagnose and optimize the pump and the works in a local computer as a master of a VFC (Variable
well. The same principle is used for PCPs, except a rotational Frequency Controller) to control the pump RPM and an
effect has to be taken into account. Of course by using the rest actuator valve controlling diluent injection at the surface. The
of the operational data that assist with construction and speed changes and injection control take place automatically
recognition of operational patterns, the PCP diagnostics as the associated instrumentation gathers all the real time data
becomes more exact and efficient for the production engineer. necessary to perform the analysis. A Host program
The work herein presented reveals field tests of the concentrates on all the wells automated, and performs
system used to optimize PCP wells, based on the important configuration, trend studies for further analysis in the
usage of axial load and several other variables (current, engineering control room.
pressures, etc) with a special ingredient that has a tremendous The system is developed using the most common
impact in optimization: Artificial Intelligence and Automation. automation and instrumentation standards, so it can be
The mathematical principles are also presented to encourage integrated to any existing automation platform. The PCPOS
the usage of axial load in a multivariable mathematical model has proven to be very reliable as it concentrates expertise and
in order to complete a good optimization scheme for these knowledge in the source code emulating an expert operator
wells. The hardware tools in the well completion needed to response by a programming technique called artificial
perform such a task are discussed as well, including the axial intelligence. The following sections include an explanation
load measuring device. about the mathematical and physical principles on which this
The benefits of operating a PCP system using a system is based. Actual field data supporting these facts are
Progressive Cavity Pump Optimization System (PCPOS) are also discussed, as well as, a brief yet informative overview of
substantial, as described in this paper. The idea is to reduce Artificial Intelligence and associated terminology.
downtime, workovers, improve system operating response
2 MENA L., KLEIN S. SPE 53962

Mathematical and Physical Principles brain works and are successfully used for pattern recognition.
It has been proven for sucker rod pumping systems that the The main benefit of neural networks is the usage of parallel
surface dynagraph chart assists with calculating downhole computing that helps fast conclusions. There are several
pump conditions. This is accomplished by correlating the axial models for neural networks: Back-Propagation, Kohonen Self-
load to the tubing and rod string forces. The wave equation Organizing Maps, Adaptive Resonance Theory from among
and the finite elements loop calculations are just several of the others, and the differences are based on complexity and usage:
methods used to do so. For PCPs, field tests performed show pattern recognition, forecasting, classification, etc.
that there is also such correlation between surface axial load As opposed to Boolean Logic (binary logic) the Fuzzy Logic
and downhole differential pressure across the pump. is a technique that emulates the human reasoning by making
This is all a consequence of the principle: “a pressure is a the variables of a problem qualitatively mean something that
force applied on a certain area” which indicates that the can be easily understood to solve the problem. The
surface axial load can be related to the forces acting on the construction and evaluation of reasoning sentences is then
pump. As a matter of fact, such relationship is the pumping easier and decisions are made in a non-restrictive
area. It is straightforward that Dynamic Fluid Level can be environment. Control rules such as: “IF the pressure IS a little
associated to differential pressure, so: bit higher than normal THEN decrease speed a little”, can be
evaluated thanks to fuzzy logic, making it easier to understand
Surface Axial Load ~ Delta P ~ Fluid Level what the conditions (values) of the variables involved mean
for a specific process.
In other words, by measuring the axial load it is possible to go Genetic Algorithms are very efficient tools for optimization
around the well-known problem of downhole measurement problems. They help determine calculations on a very wide
which is not only expensive but also, very difficult to range of possibilities for the variables of a complex problem in
maintain. order to determine the combination of the variables that
optimize the function. Remember that optimizing is
Surface Axial Load vs. Pump Differential Pressure determining maximums and/or minimums for a function
The tests compared readings of pump discharge and intake depending on the best context. For example, cost function is
pressures with downhole sensors and the axial load measured optimized when brought to a minimum and the production rate
with an axial measuring device located in the surface drive to the maximum. These functions are referred to as objective
head. The field tests performed show such a relationship exists functions.
and is considerably linear. (Fig. 1 and 2.) Figure 1, represents It is quite evident that a PCP well is such a complex process
a comparison over time of Delta P and surface axial load for that would be best solved by using advance techniques such as
different RPM. It is possible to notice oscillations along the artificial intelligence. Again the usage of all the variables
expected average value for the axial load. These oscillations involved is important for control actions. The most important
have a certain correspondence with rotational speed of the surface variables are axial load, surface tubing and casing
system and are due to mechanical vibrations caused by the pressure, current, torque, RPM, etc.
rotational motion of the system itself. Diagnostics wise, the PCPOS uses all the techniques mentioned above. The
axial load reading should be filtered and interpreted for it to variables measured in a real time mode are fuzzified and
give an adequate approximation of downhole conditions. presented to a self-organizing map that classifies and
Figure 2, represents a linear regression equation that recognizes the operational pattern. The different conditions are
demonstrate how acceptable this relationship is, the results used to evaluate a fuzzy logic inference engine to determine
obtained average the 75% for the Pearson coefficient, which control actions to be taken. In this system, the objective
makes the assumption quite reasonable. functions are production rate of the well, operational costs and
run-life of the PCP equipment.
Artificial Intelligence The main advantage of using artificial intelligence for
The artificial intelligence comprises a set of techniques that PCPOS is the fact that it acts emulating the presence of a
have the objective of emulating the human behavior and permanent expert operator at the well site taking care of every
responses by means of a software program or computer single well.
hardware in order to solve a specific and commonly complex
problem. The effectiveness of some of these techniques have Description of the PCPOS
been accepted worldwide and proven as they are applied to The PCPOS system is made up of hardware and software that
everyday and industrial problems. The main objective is to use the same principle explained above to optimize the PCP
emulate with computers the human characteristics useful for well.
decision making. Washing machines, refrigerators, car brakes The PCP System Equipment. The equipment which makes
systems, are just simple examples of areas using artificial up a PCP system best suited for the application of PCPOS
intelligence nowadays. consists of: a surface drive that contains an axial load bearing
From among others: neural networks, fuzzy logic, genetic and an axial load-measuring device; a Variable Frequency
algorithms and expert systems are most popular artificial Drive (VFD) to control the speed of the surface drive motor; a
intelligence techniques. Neural Networks emulate the way the Progressive Cavity Pump (PCP); and in the case of the heavy
SPE 53962 SURFACE AXIAL LOAD BASED PROGRESSIVE CAVITY PUMP OPTIMIZATION SYSTEM 3

oil applications a surface flowline diluent injection system done fully automatically with the proper speed-changing
(Fig. 3). device.
The Surface Drive System. In all cases to date, the surface The ingredient of Artificial Intelligence in this multivariable
drives have the axial load-measuring device or load cell approach gives a much more reliable way to control the PCP
integral to the drive. The main shaft of the drive that contains well. Of course, every artificial intelligence system needs to
the axial load bearing is directly coupled to the shaft of a go through a complete and good training procedure to generate
gearmotor. A VFD is used to vary the speed of the motor and the knowledge base to be used under operating conditions
therefore the speed of the rod string and rotor. allowing for the system to adapt to the well.
The PCP. All of the applications of PCPOS to date
consisted of various sizes (displacement and head) of PCP’s of Next is a list of achievements with PCPOS:
the 1:2 rotor to stator lobe configuration from a variety of Results Achieved:
manufacturers. Both conventional and continuous rod strings Statistical error: +5.17% compared to the down-hole
have been utilized. Sensors
The Well Head Instrumentation. This is the instrumentation -19.6% compared to the sonic fluid
necessary to supply readings to the expert system’s level measurements
mathematical model. The axial load measurement device, the 10% increase in production
tubing and casing pressure sensors, and surface fluid Increase in the PCP system life: 10 months average to
temperature sensor are mounted in the appropriate locations at Over 18 months
the well head. If required, a diluent or chemical injection Protection against two unexpected diluent plant shut
system can be located in the most appropriate location to the downs
application. The current (torque) is taken from the VFD. The Field Operators assistance cost reduction: $19,700/yr
relationships analyzed by PCPOS using these analog values Reduction in rig requirements: $3,030/yr
establish the dynamic operating conditions of the system and Reduction in replacement equipment costs: $6,725/yr
the dynamic fluid level of the well.
The Field Hardware. The field computer or RCU (Remote Conclusions and Recommendations
Control Unit) communicates with the VFD to control the -Axial load can and should be used to diagnose PCP systems
speed of the drive motor and to monitor power usage by the with the correct dynagraphic tool since pattern recognition can
system. The RCU is configured to store data and to operate the be performed.
PCP system based on basic information downloaded from the - Mechanical noise due to vibrations (caused by rotational
Artificial Intelligence Operation portion of the Expert System. dynamics of the PCP) is present in the actual surface axial
The Expert System Software. This is a C++ program load measurement, and further study of this should be carried
developed using Artificial Intelligence techniques such as out.
Fuzzy Logic, Neural Networks, Genetic Algorithms and - PCPOS yields accurate results and is generally more reliable
expert systems. The portion of the Expert System that is at or and less expensive than down-hole pressure sensor systems.
near the well location performs the advanced control rules to - A PCP well has a mathematical multivariable model that can
optimize the well production and performs advanced analysis be effectively solved and analyzed by using Artificial
and diagnostics based on the real-time input data. It generates Intelligence techniques.
the control actions over the well and alarm messages on an -A truly complete optimization system must contain an Expert
operator’s console (with detailed explanations of the alarm). System program to analyze the raw data input and make the
The Host Expert System is used as an analytical tool to necessary control action decisions without human interface.
develop mathematical models to evaluate the well Inflow - The training phase of every artificial intelligence program is
Performance Ratio (IPR) and Outflow Performance Ratio very important to achieve accurate and reliable performance.
(OPR), perform individual well historical statistical analysis,
as well as, size the optimum PCP equipment for the known References
well conditions. It is also the man-machine interface between 1. Quijada, E., Brunings, C., Mena L., Klein S.T.: “Automated
the operator and the PCP pumping system. Diagnostic of Progressive Cavity Pumps,” manuscript No. 067
presented at the 7th UNITAR International Conference, Beijing,
China, October 27-31, 1998.
Optimization of the PCP well
2. Mena L. et. Al.: “Process Intelligent Modeling through the API
Undesired operational conditions such as parted rods, near- concept for real time applications”. Paper #94-475, presented at
by-the-well flowline leaks, pump overpressure, flowline the Instruments Society of America International Conference,
overpressure, pump-off, pump-cavitation, and many others can Anaheim, USA, October 23-28, 1994
be avoided or diagnosed by PCPOS. This results in better run- 3. Giarratano; Riley.: “Expert Systems: principles and programming
life for equipment and reduction in rig and down time as well ”. ISBN 0-534-93746-6, 2nd edition, PWS Publishing Company,
as operational labor. On a real time basis, the control rules are 1993.
evaluated in order to recognize operational conditions 4. Rich E.; Knight, K.: “Artificial Intelligence”, 2nd edition, ISBN:
different to “Normal” and mapped to a decision-making 84-481-1858-8, Mc Graw-Hill, 1994.
system to generate the associated control actions. This can be
4 MENA L., KLEIN S. SPE 53962

5. Klein S.T.; Mena L.,: “Well Optimization Package for

Progressive Cavity Pumping Systems,” SPE paper No. 52162
presented at the SPE Mid-Continent Operations Symposium,
Tulsa, OK, March 28-31, 1999.
SPE 53962 SURFACE AXIAL LOAD BASED PROGRESSIVE CAVITY PUMP OPTIMIZATION SYSTEM 5

Axial Load, Delta P vs. RPM and Time

16000 700

14000
600

12000
500

10000
Axial Load
400
Axial Load
8000 Delta P
Delta P RPM
300
6000

200
4000
RPM
100
2000

0 0
Ho 8: 8: 9: 9: 9: 9: 9: 10 10 10 10 10 11 11 11 11 11 12 12 12 12 12 13 13 13 13 13 14 14 14 14
ra 44 56 08 20 32 44 56 :0 :2 :3 :4 :5 :0 :2 :3 :4 :5 :0 :2 :3 :4 :5 :0 :2 :3 :4 :5 :0 :2 :3 :4
8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4
Time

Fig. 1 Actual Field Data comparison between Axial load

Delta P and RPM

Linear Regresion Axial vs. Delta P

15500

14500

13500

12500 y = 8.1875x + 7132.3

Axial Load vs. Delta P
2
11500 R = 0.7881 Linear (Axial Load vs. Delta P)
Linear (Axial Load vs. Delta P)
Linear (Axial Load vs. Delta P)

10500

9500

8500

7500
200 250 300 350 400 450 500 550 600 650

Fig. 2 Linear Regression between Delta P and Surface Axial Load

6 MENA L., KLEIN S. SPE 53962

Fig. 3 Typical PCPOS Equipment Installation