Challenging the current barriers to forecast improvement
There are several practical constraints currently facing the industry in terms of its ability to improve production forecasting.
Manpower constraints (quantity, available hours)
This will only get worse in the coming decades (big crew change, PE university enrollment, etc.)
Manpower constraints (quality, i.e. level of training)
Note that the academia have an important role here to prepare engineers etc. for their future jobs. In my opinion, they fail in terms of bringing the future PTPs (Petrotechnical Professionals) into the ‘correct’ mindset, i.e. how to deal with an ill-determined system, how to use probabilistic methods, how to classify the various types of uncertainty + their relation to modelling, how to ‘exploit’ uncertainty: how to link uncertainty to decision-making, how to learn formally from the response of an ill-determined system, how to generalize such learnings, how to mature the evolving knowledge in terms of risk mitigation and upside capturing, options thinking, how to distinguish assumptions/controllables/non-controllables, how to set/use targets, how to propagate uncertainty through complex models, how to consolidate uncertainty over multi-assets, how to reward staff (on outcome vis-à-vis ‘target’?, or on how they intelligently used / improved PF workflows), how to define ‘robustness’ of a forecast, etc. etc.
Most PTPs are essentially trained in a deterministic mindset. This negatively affects their later chances of putting themselves in the probabilistic mindset, which is essential, as our subsurface in essence is an ill-determined system (unlike other engineering disciplines). Note that quite a bit is being done by the academia, and later at the industry training courses, but in my opinion, by far not enough, and too late. I would label this ill-preparedness of the future PTPs as an intellectual constraint: later in their careers they have big problems grasping the essence of this mindset, and (continue to) operate in a mindset that results in avoidable biases, harming the industry.
Lack of agreed best practices (methodological constraints)
Notably, no definition of an ideal practice (including formalized learning)
And notably, a lack of controlled experiments that convincingly demonstrate how cutting methodological corners yields bias and incompleteness + what is acceptable (good enough / fit-for-purpose).
Value proposition constraints
There may be a degree of short-termism that potentially negatively affects the value of improved forecasting methods: medium-term and long-terms forecasting (which are much more susceptible to reservoir parameter uncertainty + applicable physics uncertainty) receive less attention (i.e. in terms of grasping methodological improvement opportunities) than short-term forecasting (which is mainly a function of operational variables and, hence, less subject to parametric and physics uncertainty).
This is exacerbated by the practice of compounding disparate sources of uncertainty and mixing them with controllable parameters: setting targets such that they can be met by changing the assumptions (controllables) of the original forecast, bereaves the forecasters from the opportunity to learn methodically.
Obviously, uncertainty and design flexibility are very much related (options thinking). I guess more ought to be done in terms of options valuation to demonstrate the added value of flexibility options that allow the operator to steer the operations in a more profitable way as new information is being revealed in time.
Organizational (competitive) constraints
Notably, a lack of international cooperation to advance science and best practice. This is probably due to the competitive model in the industry: there are no governments forcing a better international collaboration (such as the international meteorological societies who can demonstrate that their forecasts have improved over the years: they have turned forecasting into a science).
Hence, the E&P industry has no culture of collaborating in (pre-competitive) R&D. For example, sharing data in an open and transparent way still is a big hurdle. Obtaining field data & models with earlier forecasts vs. actual performance is problematic (confidential).
Practical tool constraints
- No articulated vision on an ideal practice, let alone on how a tool environment should look like that enables such ideal practice.
- Note that with the big crew change, and the loss of valuable experience in the industry, improved software and knowledge systems should have an even greater potential value. If only to avoid some of the big avoidable mistakes that will undoubtedly be made in the future.
Perhaps there are other classes of constraints. The above is somewhat different from the standard time-budget-manpower classification. I believe that, for improved production forecasting, budgets should not be a constraint: provided a convincing proposition would be made to improve forecasting, then the budgets will be made available. But the problem is in convincing the people who would have to approve such budgets: they seem mostly happy to use existing methods and tools. They are not convinced that the production forecasting state-of-the-art should be / can be improved. The looming E&P manpower bottlenecks is not seen as a threat requiring improved (more formalized) production forecasting methods.