Oil and Gas Decision Making - Why we keep underplaying uncertainty ?




As the oil industry has high challenges running day to day activities such as reservoir monitoring, management of the field, field development planning as well as estimating reserves and resources. Deterministic workflows reflect the capacity that we have as humans to process information, but the nature of our work should be stochastic to account for uncertainty. This article will be the first in a series of articles explaining the current industry environment towards uncertainty analysis and how as petroleum engineers we can efficiently make use of statistical learning techniques to help us in the movement from deterministic to probabilistic workflows





Figure 1 The analysis process (Model,2016)

With the increase in computational power (Figure 2) we are now capable of running multiple scenarios utilizing several tools from very basic analytical tools such as material balance and decline curve analysis in order to conduct thousands of simulations in a matter of seconds. On top of this we can propagate uncertainty using these tools.



Figure 2 Computational power over the years Max Roser (2016)

We see this advantage of an increase in computational power being utilized by companies which develop reservoir simulators (Figure 3). We therefore find ourselves as petroleum engineers with the opportunity to leverage and also take advantage of this increase in computational power not only in understanding the physics of the reservoir for example, but also in the added value of understanding the uncertainty allowing us to really explain why we believe that this is the potential outcome of this particular task.


Figure 3 Most common reservoir simulators


We have come across projects where we have been asked about our evaluation of the best technical case or your P50 estimate as well questions such as how you would protect against a potential low outcome such as a P90. So depending on the tools that are available in order to model the system, subsurface and surface, we will have to evaluate whether we use deterministic up to fully probabilistic techniques. Through these series I will introduce uncertainty analysis in terms of the decision making process, what effects it, introduce probability distributions and their use in generating Monte Carlo simulations as well as how to be more efficient by using experimental design.

So what is decision making? It can be defined as a cognitive process that results in the selection of a sequence of actions among other sequence of actions. Here we have the word cognitive which comes from cognition which just means the process of acquiring and understanding knowledge through our senses, experiences and thoughts. So in simpler terms decision making is a process that utilizes our senses, experiences and thoughts in order to come up with a series of actions.

So what is the problem with decision making? The problem with this is that this can be influenced by our current and immediate environment on a sub conscious level.  So for example when we are dealing under a higher pressure or higher stakes we tend to use an intuitive decision making process as opposed to a structured approach which will neglect any alternative action there may be. So what influences the decision making process? A number of biases influence the decision making process some of which are (Figure 4): Confirmation bias: Search for evidence that’s supports certain conclusions whilst neglecting other evidence which supports other conclusion.
  • Premature termination: Accepting first option which seems right.
  • Selective perception: screening information which one may deem unimportant.
  • Recency: Paying more attention to new information and ignoring the old.
  • Credibility of source: Bias where one rejects statement on the basis of bias against another person.



Figure 4 Factors/Biases in Decision Making

 


A number of other biases exists such as the anchoring bias (Figure 5). This is when people tend to base estimates on any value at hand. A number of O&G companies try and tackle this particular bias by first asking their employees to give an estimate of their ranges before giving their best guess in an effort to prevent anchoring. The bandwagon effect is the tendency of humans to do or think things solely because other people do or think them. The blind spot bias describes the failure of recognizing one’s own biases but being able to recognize them in others. The confirmation bias describes the tendency of people to conform around the information that confirms our views.


Figure 5 Some Types of Cognitive Biases (Wood, 2015)

A number of surveys as well as published literature has looked at the topic of cognitive biases in the Oil & Gas Industry. Welsh in 2014 published a paper titled “Experience & Cognitive Biases in Oil & Gas Industry Personnel” where he studied anchoring and overconfidence bias in O&G personnel. He gave a question where he asked “What was the global reserves estimate in 2003?”.

 As a part of the question he gave an anchor by asking if the reserves were higher or lower than of 573 Billion bbl or higher and lower than 1721 Billion bbl. He showed that from 176  participants, 85 saw the lower anchor and 91 saw the higher anchor (Figure 6) . For curiosity sake, BP in 2003 had the total global reserves estimated at 1,147.7 billion barrels. He concluded by saying that those who are trained have a marginal advantage over those with no prior training.




Figure 6 Mean Estimated World Proved Reserves 2003 (Welsh, 2014)

 
Another survey titled “The Difficulty of Assessing Uncertainty” by Capen in 1976, asked SPE members to put ranges around 10 questions some of which included:


- What is the area of Canada in square miles? 
- How long is the Amazon River in miles?
The study concluded that:

- People who are uncertain about answers to a question have almost no idea of the degree of their uncertainty. 
- The more people know about a subject, the more likely they are to construct a large probability interval
- Even when people are told that their ranges are too small, they cannot bring themselves to make them wide enough (although they do somewhat better).Most don’t know what a reasonable range is. 
- Their ranges are based on what they see as a possibility but since uncertainty comes from events which we do not foresee the ranges we end up with are too narrow.

So how is the oil and gas industry preforming in quantifying uncertainty? Demirmen in 1999 found that from his survey where he asked about reserves estimation and the handling of uncertainties for oil fields in the UK found that 40% of the fields underwent a 50% reserve change over a 10 year period. He also concluded with a list of the most difficult factors to account for which were:



- Geological uncertainty
- Well Failure
- Capex/Opex
- Oil & Gas Prices
- Reserves in mature fields

A Schlumberger survey in 2015 concluded that although there was improvement in uncertainty quantification, it did not reflect in improved decision making.  A 2015 survey also conducted by Schlumberger reflected that 80% of O&G professionals accounted for uncertainties but only 37% felt that they were well understood.

A recent survey conducted by Primera Reservoir in conjunction with Portsmouth University proposed a number of questions to petroleum engineers, the three main questions being:

- How do you perceive the values of probabilistic workflows?
- How would you best describe the approach mostly used for forecasting? 
- What is the proportion of projects that have a probabilistic approach within your company?


The results of the survey are portrayed below (Figure 7), showing that just above 50% of the participants had a high perception of probabilistic workflows.  Almost 60% of the participants used a deterministic approach when forecasting and almost 40% said that only 20% of the projects within their company took on a probabilistic approach. Only 5% said that always took a probabilistic approach. This is an indication that as an industry we have yet to come to terms with how viable probabilistic methods can be for uncertainty analysis and quantification.


Figure 7 Primera Reservoir & University of Portsmouth (2014)

So what is the Oil & Gas Industry approach when dealing with uncertainty analysis? For the most part we see that the exploration side has taking the probabilistic (use full range of values that could possibly occur for each unknown parameter, generating a full range of possible outcomes) approach as being the norm whilst in the production side a deterministic (using a single values for each parameter, resulting in a single value e.g. Best Technical Estimate) approach is taken.



Figure 8 Some of the decisions faced during life of field (The Digital Oil Field, 2004)

Some examples of the decisions that O&G professionals may face during the life stages of a field are portrayed above (Figure 8). For example the decision to invest capital has uncertainty regarding:

- The recovery of the field
- The production profile
- Number of wells
- Facilities required


This could result in risks with outcomes such as:

- Delays
- Cost overruns
- Inability to deliver
 


This gives an indication of why we need to conduct uncertainty analysis, so the question now becomes what are the tools that we have in order to quantify these risks. A workflow explaining this is presented below (Figure 9).  Of course uncertainty analysis is not just specific to the examples illustrated above but also those that are presented in this workflow such as those uncertainties associated with reserves, infill drilling and reservoir management.



Figure 9 Why, What & How in quantifying the risk in oil & gas

In terms of the problem solving tool or what we use to quantify the “Why”, we have material balance, decline curve analysis, reservoir simulation or integrated asset modelling. How we quantify this is either through deterministic, deterministic and stochastic, ranking of geological models mixed with fully stochastic dynamics or fully stochastic involving Monte Carlo applications.

All of this can lead up to the economic analysis which has taken a deterministic norm in the industry. This leads up to sensitivity analysis, propagation of uncertainty (key performance indicators), running regressions and creating proxies.

In the following articles I will look into introducing how we can use these tools in order to help us in propagating our uncertainty.
 
Author: Ahmed Muftah 
Petroleum Engineering Consultant 
SDM, CEOR Specialist
 
References

·    Max Roser (2016) – ‘Technological Progress’. Published online at OurWorldInData.org. Retrieved from: https://ourworldindata.org/technological-progress/ [Online Resource]

·     Welsh, M.B. (2014) Experience and cognitive biases in oil and gas industry personnel. Available at: http://www.psychology.adelaide.edu.au/cognition/aml/aml2/welsh_aml2.pdf (Accessed: 6 February 2017).

·    Wood, J.M. (2015) 20 cognitive biases that affect your decisions. Available at: http://mentalfloss.com/article/68705/20-cognitive-biases-affect-your-decisions (Accessed: 6 February 2017).

     Capen, E.C. and Co, A.R. (1976) ‘The difficulty of assessing uncertainty (includes associated papers 6422 and 6423 and 6424 and 6425 )’, Journal of Petroleum Technology, 28(08), pp. 843–850. doi: 10.2118/5579-PA.

·    Demirmen, F. (2009) Distinguished author series - reserves estimation: The challenge for the industry. Available at: http://large.stanford.edu/courses/2013/ph240/zaydullin2/docs/demirmen.pdf (Accessed: 17 February 2017).



0 comments: