SPE 165842

Achieving Production Optimization Using Progressive Cavity Pumps,

Artificial Neural Networks, and System-Based Monitoring
J uan Eggers, SPE, Schlumberger; Angel E. Gomez, and Yomalys Hurtado, Petrocedeo; Amin Claib, SPE,
Petrocedeo; Tayruma Silva, SPE, and Gustavo J . Nunez, SPE, Schlumberger; and J esus Borjas, Luis Duarte,
and Sinaira Valbuena, Schlumberger
Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Heavy Oil Conference Canada held in Calgary, Alberta, Canada, 1113 J une 2013.

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 have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is 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 acknowledgment of SPE copyright.


Abstract

Petrocedeo is a Venezuelan joint venture (JV) between PDVSA, Total, and Statoil. Petrocedeo operates in the San Diego
field of the Junn block located in the Orinoco Belt, where over 500 lifted wells currently produce over 120,000 bpd of extra-
heavy oil (7-9.5 API) using progressive cavity pumps (PCPs). Petrocedeo has extensively used down-hole sensors to
monitor PCP operational parameters, such as velocity, torque, vibration, intake and discharge pressure, and temperature.
Timely and proper usage of this data, incorporated with other well information and operational data, improves production
optimization using proactive surveillance and diagnostics of underperforming wells. In spite of having process data in hand,
it was noticed that additional optimization can be obtained by integrating the data available with articulated workflows.
Initially, a pilot was conducted to evaluate technology aimed at optimizing the production of 50 wells through timely
identification of underperformance occurrences. This was achieved using:
Automated data gathering and integration
Automated daily production rate estimation using operational data and artificial neural networks (ANNs)
Customized surveillance and diagnostic workflows
The technology applied was developed by integrating data and estimating well production rates on an hourly basis. This
involved using trained ANNs and a leading production technology platform. In addition, continuous surveillance workflows
of operational parameters as well as estimated rates and other production information were implemented on engineers
computers through customized well and reservoir analysis software created during the pilot.
After the pilot project, Petrocedeo engineers were able to reduce the time to identify underperforming wells in 20%. The
positive results achieved in the mentioned pilot, encouraged the company to implement the system in the whole San Diego
field, as well as introducing additional production surveillance and optimization workflows and visualization tools. This
paper presents some of the main workflows implemented and the results obtained.

2 SPE 165482


Introduction

Petrocedeo has faced significant challenges over the last few years to meet production targets. After determining that
production losses were mainly related to reactive surveillance, Petrocedeo looked for production maximization based on
proactive surveillance and diagnostics of underperforming wells that could contribute to meeting the production goals, while
maximizing recovery factor and investment return.
As many other companies, the key challenge that Petrocedeo face to proactively identify underperforming wells is the
limitation of carrying out production well tests that are frequent enough. To overcome this, the authors proposed an
innovative scheme that estimates well production rates by using trained artificial neural networks (ANNs) fed with
operational parameters. Workflows based on these estimated rates in conjunction with operational parameters from sensors
allow identifying and diagnosing underperforming wells.
To prove the concept Petrocedeo performed a pilot implementation that will be shown on the following pages. The pilot
project was developed using a sample of 50 wells.

Production Surveillance Challenges

When Petrocedeo was formed in 2007, they received a highly instrumented field from the previous operator. Petrocedeo
have preserved the instrumentation, data transmission, processing and monitoring tools since then. Even though these tools
proved helpful in operating and monitoring San Diego field for several years, the absence of an efficient and standardized
process for field surveillance as well as underperforming wells detection and diagnostic is preventing Petrocedeo from
achieving production targets now that the field has over 500 producing wells, a thirteen year production history and is
considerably depleted.
Underperforming well identification was based in operational measurements and inferences, mainly from pumps data.
Production well rates were unknown for several weeks; as well tests are in the best case carried out once a month, providing a
direct measurement of wells deliverability. In heavy oil wells, overall system optimization is a continuous process that
requires real time data to be analyzed in order to issue recommendations about the optimal operational settings that apply at a
given time, like the diluent injection rate to mention one example.
Wells surveillance was performed by looking at summaries of most important operational parameters that are fed from the
multiple data sources into several spreadsheets. Operations, Optimization and Reservoir Engineers performed a well by well
review using such spreadsheets on a frequency that was limited by the number of wells assigned to each and data loading
speed. Pump and other well issues detection using primarily spreadsheets was one of the most time-consuming activities that
engineers performed on a daily basis. Duplicity of values for the same parameters, slow performance, lack of adaptability and
maintainability from this old technique had become a great challenge to production and optimization engineers.
As a consequence, pump failure rate, well intervention frequency and deferred production have substantially increased.
In a nutshell, Petrocedeo were facing the challenge of achieving production targets without an effective way to identify
production losses, well and pump issues in a timely manner.

Software Solution Overview

Aiming to facilitate asset optimization in the long term, a multi stage project was conceived, where basic steps were set as the
priority to develop a complete solution. Foundation for this solution is based on the pyramidal scheme presented in Figure 1.

SPE 165482 3




Figure 1. Stages of a Production Optimization Project

At the initial project stage, the goal was building the first two levels of the pyramid: data management, and monitoring and
surveillance, as well as a simple analysis for detecting underperforming wells. This had to be performed while maximizing
the use of Petrocedeos existing databases and software base. This called for the definition of three main layers for the
solution to be implemented, which are shown in Figure 2.

Data Management
Monitoring
& Surveillance
Surveillance Visualization & Alarming
Data Acquisition & Conditioning
Operations & Volumes Data Management
Allocations & Reporting
Analysis
& Diagnostics
Engineering Diagnostics & Analysis
Well and Surface Network Design & Modelling
Well, Network & Asset Optimization
Field Development Planning
Optimization
4 SPE 165482

Figure 2. Production Optimization Solution Layers

Data integration layer.

This is the core of the solution, it collects and gathers all data available from different sources and sets automated data
processes for routine time consuming data acquisition and data handling activities performed by engineers. The main purpose
of this layer is to integrate and enable all data in a single data repository, while minimizing data bases manual updating by
automating data acquisition processes.
The data bases that Petrocedeo used for production optimization purposes were:
Drilling and Well Interventions data base: it has all data from drilling operations, well interventions and pump
replacements.
Production data base: mainly used for production allocation and storage of well tests, fluid samples, well events
chronology, and other types of production information.
Historian: SCADA data historian; this is the database of real time operational data.
Spreadsheets: they had been widely used in Petrocedeo for well surveillance and collecting all sorts of information
such as well potential, pump running days, pump parameters, amongst others.
Fluids sampling data base: it gathers lab test results for wells fluid sampling performed.
In this legacy scheme, the duplication of data was caused by single purpose development of spreadsheets from all groups:
reservoir engineers, production and optimization engineers, field operations and well engineers group. Each working group
replicated as much data from different data sources as required by a single purpose. Interaction between data sources and the
working groups is summarized in Figure 3.
SPE 165482 5

Figure 3. Legacy relationship between production optimization data sources and spreadsheets used by working groups

In order to simplify and avoid data duplicity, a simple new scheme was proposed by Schlumberger, based in a single data
integrator where all kind of data can be centralized and stored, from real time data to daily production, well events and other
types of all frequent, sporadic, static and dynamic data.
From this layer a complete data base was created, including historical available data for each well as well as automated
data acquisition for the following parameters:
Static parameters: such as general well, location and reservoir information.
Daily and Sporadic data: pump replacements, production allocations, well testing, diluent injection point, fiscal
production rates, fluid sampling results, well potential, amongst others.
Real-time data; intraday data from down-hole and surface sensors of well head temperature, tubing head pressure,
casing head pressure, pump discharge pressure, pump intake pressure, pump intake temperature, pump torque, pump velocity,
diluent injection rate, pump discharge temperature, diluent injection pressure, pump vibration (x,y and z), amongst others.

Workflow Integration layer.
This layer was developed to maximize value by estimating well production rates from operational data. On this layer, field
measured data is processed for further analysis. Automated daily production rate estimation by using operational data
available and Artificial Neural Networks (ANNs) is performed on an hourly basis.
The methodology used was:
Identification of PCP models used on the 50 wells selected for the pilot, for a total of 18 PCP models
Creation of a data set of pump velocity, pump differential pressure and production rate from theoretical PCP
performance curves.
6 SPE 165482
Creation of neural network training patterns (test, training) for the 18 PCP types
Design of ANNs topology and training from patterns. ANNs were designed for each of the 18 pump models
installed on the 50 wells. They calculate estimated liquid rate from pump differential pressure and velocity.
Using the ANNs, well test and well potential information, two key performance indicators (KPIs) were calculated based
on the two workflows explained next.

Pump performance factor workflow.

The pump performance factor is a KPI calculated every time that a well test is performed. It is defined as the ratio of the
production rate measured during the well test and the theoretical rate the pump can deliver at factory conditions, for the same
pump velocity and differential pressure values:

PPF =
Qwtest
QANN(RPN, P)wtest


where:
PPF = Pump Performance Factor
Qwtest = Well Test Production Rate
QANN(RPN, P)wtest = Production Rate calculated by the ANN at well test pump speed (RPM) and pump pressure
delta

Liquid rate calculated by the ANN (QANN) is calculated every time a well has a new well test; using the pump velocity
and pump differential pressure existent during the well test performed.
The PPF is later used in other workflows for adjusting production rates estimated by ANNs.

Well performance factor workflow.

This workflow was set to be performed on an hourly basis. Its primary function is help identifying underperforming wells
in a timely manner:

wPFt = i =
cst t = i
potcnciol t = i


cst t = i = 24
cum t=

cum t = i = t = i 24
t=
t=1


t = i = ANN t = i PPFt = i

where:
WPFt=i = Well Performance Factor at time i
Qfcst t=i = Daily Production Rate forecasted at time i
Qpotential t=i = Current Well Potential
Qcum t=i = Todays cumulative produced volume until time i
i = number of hours elapsed today
SPE 165482 7
Q t=i = Estimated Daily Production rate at time i
QANN t=i = Production Rate calculated by the ANN at time i
PPF t =i = Pump Performance Factor at time i

The first step that WPF workflow performs is an instantaneous liquid rate calculation (Q t=i). It is calculated using the
ANN at the pump velocity and differential pressure gathered from down-hole sensors for time i. If for any reason this data
is not available at the time the workflow is run, other complementary workflows are triggered, which estimate down-hole
pressures using other measured field data.

Then a cumulative daily production is calculated from hourly instantaneous rates sum. This cumulative production is in
turn used for forecasting the estimated volume that the well will produce during the complete day, as shown in Figure 4.












Figure 4. Visualization of Estimated Dail y Rate calculation

Once the full day estimated volume is calculated, it is divided by the well potential to evaluate the well performance.
When an event that affects the well performance happens, an unusual behavior is observed, as shown in Figure 5.

8 SPE 165482


Figure 5. Visualization of Estimated Dail y Rate calculation for a well witnessing a production drop


Visual integration and collaboration layer.

At this layer, daily monitoring and surveillance activities take place by efficiently displaying all sorts of field information
collected through the existing systems, in addition to well production rates and key performance indicators calculated in the
workflow integration layer. This environment enables engineers to perform efficient analysis of current field, wells and
pumps performance, allowing fast and well supported decision making, with the ultimate goal of reducing pump failures,
well issues in general and deferred production, as well as identifying opportunities for production optimization.
The data is integrated into surveillance-by-exception engineering processes that avoid repetitive manual tasks. Each
engineering process consists of a predefined sequence of tasks that narrow the large amount of existent data, targeting the
efficient identification of potential failures in PCPs and allowing engineers to identify abnormal well behaviors. The
objective is to have an environment where the user has sufficient time and information to make good decisions quickly.
The workflows combine sporadic and continuous data streams from data historians, spreadsheets and corporative
databases. The variety in source data frequency requires a preliminary analysis that confirms a low sensitivity of the results
on the assumption that sporadic properties remain constant between successive samplings.
During the initial phase of the project, a very flexible well and reservoir surveillance software platform was used to
visualize and create all the engineering workflows required. This included the creation of multiple plots, reports, customized
filters, queries, maps and other analysis tools at different levels, which could later be adjusted and improved according to user
needs.
Some of the tools developed that have a high impact on the daily operational surveillance tasks are explained below.

Base Map.

SPE 165482 9
The Base Map is the initial view for the users. It allows identifying the wells according to different parameters, static or
dynamic, measured or calculated. The user can explore trough the different panes and select the workflows desired.



Figure 6. Field Surveillance Base Map


Scatter Plots.

Macro visualization of all wells in single plots that facilitate detection of underperforming wells was achieved by creating
scatter plots of diverse operational parameters in conjunction with KPIs. Some of them can be observed in Figures 7, 8, 9 and
10.




Figure 7. Pump Intake Pressure vs. Pump Performance Factor Scatter Plot
10 SPE 165482







Figure 8. Pump Performance Factor vs. Liquid Rate Scatter Plot




Figure 9. Torque Ratio vs. Intake Pressure Scatter Plot



SPE 165482 11


Figure 10. Pump Vibration vs. Pump Velocity Scatter Plot


Pump Monitor - Well Information & KPIs.

Daily activities of optimization engineers include but are not limited to performing surveillance tasks for early detection
of failures. The Pump Monitor plot assists them in performing that task by showing current and historical PCP operational
conditions. An example of it is included in Figure 11. It is divided in three sections: surface conditions, subsurface conditions
and PCP panel parameters. Additional to the plot, the Well Info pane (left) shows pump key performance indicators and
last well test data.




Figure 11. Pump Monitor Screen


Additional plot templates were created in order to facilitate analyses and diagnostic work performed by reservoir,
production and optimization engineers. These graphs are defined to help visualize pump parameters, key performance
indicators, well test data and general production information.

12 SPE 165482


Solution extension.
Following the pilot described, and encouraged by the positive results, Petrocedeo decided to extend the surveillance
system to the whole San Diego field, which currently has about 500 active wells. In addition to this, the functionalities of the
system were expanded, to include additional surveillance and diagnostic workflows, visualization and collaboration tools, as
well as incorporating some key surface production process equipment. This improved surveillance and diagnostics system
reaches the third level of the pyramidal scheme presented in Figure 1: Analysis and Diagnostics.
One of the main features of it is the web visualization developed during the second phase of the production optimization
project, which is now allowing Petrocedeo to monitor the current field situation from anywhere at all times, presenting
highly customized views of measured and calculated operational parameters, performance indicators and a vast amount of
historical and real time information for all wells, PCPs, key process equipment and the field in general. A comprehensive set
of well, pump and cluster status KPIs was also developed, which allow engineers to quickly spot the areas that require
attention at a given time and prioritize their work and actions accordingly, saving a huge amount of surveillance, analysis,
diagnostic and decision making man hours, and more importantly, reducing issues detection time with the consequent
reduction of well, pump and equipment failures and their corresponding production losses.
This solution extension implemented in the second phase of the project will be explained in much more detail in a
separate paper. Some of the web accessed screens developed are presented in Figures 12, 13, 14 and 15.


Figure 12. Main Field Level Web Surveillance Screen

SPE 165482 13

Figure 13. Main Cluster Level Web Surveillance Screen


Figure 14. Main Well Level Web Surveillance Screen


Figure 15. Main Management Web Surveillance Screen

14 SPE 165482

Results

A completed implementation on 50 wells of structured surveillance workflows and system based monitoring with KPIs
defined for Petrocedeo was delivered during the initial phase of the long term production optimization project.
Petrocedeo engineers were able to reduce the time to identify underperforming wells in 20%. This was achieved by a
simple drilldown troubleshooting that started using scatters plots to limit the number of wells to the group of wells with
unusual operational parameters. This quick surveillance screening performed using the developed plots and reports in the
single visualization platform, enabled engineers to easily detect underperforming and potential issue wells, as well as the
likely cause of underperformance or issue.
This pilot implementation established the foundation for a complete optimization structure and enabled the establishment of
standardized procedures for well monitoring, surveillance, analysis and diagnostics based on Key Performance Indicators.
The second phase of the project is being completed at the time that this paper is being written, so the results of it will be
presented in a future paper. A significant amount of positive results are expected from it, most of them related to engineering
time savings, production issue detection time decrease, pump failure prevention and failure rate reduction, and deferred
production decrease.

Conclusions

The capability of Artificial Neural Networks using operational parameters measured in real time for estimating daily
production rates of wells producing with Progressive Cavity Pumps was proved during the pilot project performed.
Beyond that, the inclusion of that ANN estimated well rate in articulated-automatic-surveillance and analysis workflows,
allowed Petrocedeo to have a more comprehensive understanding of how the wells, pumps and the field in general are
performing at a given time.
Furthermore, the integration of this estimated rates with a vast amount of other KPIs and production information coming
from multiple sources and of different nature, like frequent and sporadic, down-hole and surface, measured and calculated,
has facilitated the surveillance and analysis work enormously.
Finally, the standardization of field monitoring, surveillance and analysis work, as well as the centralization of production
optimization most relevant information in a single platform, are additional benefits that have been possible through this
system-based monitoring project.
All these benefits have already helped reduce wells failure detection time and saved man hours, and will eventually impact
pump failure rates and deferred production, contributing to achieving corporate production objectives.
The success of this pilot phase led Petrocedeo management to approve project continuation to the second phase, which
involved the extension of the system from 50 to the whole 500 active wells of San Diego field, as well as an expansion of the
functionalities to include additional surveillance and diagnostic workflows, visualization and collaboration tools, as well as
incorporating some key surface production process equipment.
Additional benefits will be realized once the second phase of the long term production optimization project is fully
operational.

SPE 165482 15
Nomenclature

ANN = Artificial Neural Network
KPI = Key Performance Indicator
PCP = Progressive Cavity Pump
PPF = Pump Performance Factor
RPM = Revolutions Per Minute (Pump Velocity)
SCADA = Supervisory Control And Data Acquisition system
WPF = Well Performance Factor.