| dc.description.abstract | This thesis is performed for Statoil in Statoil premises in Stavanger. Statoil is a large offshore and subsea
oil and gas operator in Norway, with increasing activity internationally, both offshore as well as on land. The
thesis will look at Statoil’s experience data, mainly in Norway, and compare these with established learning
curve theory in the drilling and well business, in order to develop an updated method for learning curves in
the time estimation process in the company. Time and cost estimation is tightly linked and this thesis’ main
contribution will be to the time aspect. The time and cost aspects of drilling and well operations have gained
increased attention during recent years, with strongly escalating cost, and the thesis is a part of the initiative
to improve the estimation process.
There is in the order of 100 wells drilled and completed each year in Norway, 70% of these by Statoil.
Statoil is also involved in operations internationally, in example South America, Iraq, West Africa, Gulf of
Mexico and Shale Gas in USA. The cost of these activities is considerable, hence the importance to
maintain effectiveness, and even better; “to improve performance both short term and over time”.
This report is based on data acquired from 97 drilling facilities and 3267 wells. Ideally the first wells should
have a steep learning curve, towards the technical limit, where the minimum time usage is reached for the
lifetime of the investment.
Earlier findings indicate that there are different opinions and understandings within Statoil of what learning
curves are, as well as what they should be used for. The main application of learning curves is time
estimation. To verify or improve the current learning curve used in drilling, Completion and Well Operations
today, DBR data have been used for learning curve analysis. The analysis has been split into different rig
types. A Learning Curve Analysis tool has been developed in Excel, where DBR data is imported and
analyzed for the proposed new learning curve parameters and estimation process. The new learning curve
parameters also include a delay parameter in addition to the standard Brett Millheim parameters.
Results indicate a poor level of learning measured by the Brett Millheim C2 learning coefficient and alfa
value. The alfa value tells how much additional time is added to the technical limit, in relative terms. There
is also higher spread in the semi submersible results compared to the others rig types. The larger spread
gives very high max values for the lowest efficiency wells. This is suggested to be an area of increased
focus, both because of the performance numbers, but also because a very large portion of Statoil’s wells
are drilled with semi submersible’s, and last because the rig rates usually are the highest for these rigs. And
hence even small improvements can give large improvements.
Based on the results from the analysis and specific findings, some rigs were selected for a more in-depth
analysis. Good examples with good field related learning have been found when using an old semisubmersible
drilling rig for a series of comparable wells. A brand new rig drilling with its first wells is also
presented.
Another question is also how improved and faster learning can add the most value to the company? This is
a question with many different opinions, however high involvement, in Statoil today. This thesis would like to
promote the “knowledge and value approach”, meaning that increased understanding of time usage and
learning processes can become a useful (and powerful) tool to visualize where on the learning curve a
specific well are located, in order to be able and select reference wells more knowledgeable. And in
operation, help to see in what direction the efficiency are going. It is suggested that the personnel
III
influencing the “underlying large cost drivers” such as method selection, procedures, etc. should have easy
access to learning curve competency and information.
The potential upside from an increased focus on learning curves as a management and operation tool is
considerable. The economical significance is in the in the order of being able to justify projects or not.
Numbers like 25, 35 and even 50% and higher reduction in time usage is mentioned in literature, if you
manage to raise the learning level towards a normal and excellent level.
It is suggested to:
Improve the data quality in DBR for key parameters used in LCA analysis
Include Learning curve analysis in the Drilling and Well Estimator (DWE)
Perform Learning Curve Analysis (LCA) follow-up on completion and Well operations
Continue and expand relationship with University of Stavanger (UiS) and Petroleum and Asset
Management departments (Aadnøy and Liyanage)
Platforms where the drilling facilities have been upgraded to meet new HSE standards, etc. have been
looked upon in order to see if possible to measure performance before and after with respect to learning.
These upgrades usually takes place late in the field life, and the drilling targets before and after upgrade are
very scattered as well as very few completely new wells drilled (one of the analysis criteria). Instead one
finds a lot of different types of sidetrack’s and some extended reach wells making it difficult to compare
directly before and after upgrade. It is suggested to go one step more into detail in the dataset in order to
compare time for parts of the drilling process, similar to what has been done in the pilot study of this thesis.
It is also suggested to include the LCA analysis tool into the next version of the DWE estimator, where a
natural framework is already built-in. An LCA analysis ability applied here will improve the understanding for
what learning curves represent in practice, and what factors have an influence.
In the data made available there has been found for the existence of a “forgetting factor”. This is further
described in this report. There has, also, been found support in the data to claim “learning delay” in some
types of operations. The thesis work has included the following sub-processes:
Literature study, to find out about and verify the industry standards.
Set-up measurement parameters.
Develop tools and procedures for data extraction from the drilling reporting database (DBR) and
Learning Curve Analysis (LCA).
Perform in depth analysis of 97 drilling facilities and 3267 (qualified) wells from the DBR data base
system. A total of 5-10 000manual data steps have been performed
Establish guidelines for dealing with uncertainties and finding the limitations in the dataset.
Main focus has been startup of drilling facilities on fixed platforms, and the first (5) wells.
Assessment of whether the current Statoil time estimation learning curve application model is OK,
or if (the) industry standard give a better description
Analysis of whether we see a learning curve at all. Can we better predict where and when we will
see, and where we will not see a learning curve?
Can project learning curves be estimated in advance?
How we can improve learning speed has also developed as the work progressed, and maybe a
separate thesis is needed to properly answer such challenges.
Conclude with several specific focus areas for future work, both internally, and together with an
academically institution. As well as some findings, and indications that can be useful in the ongoing work
within performance measurement and management towards an even more competitive Statoil in the future. | no_NO |
