Eight years to find an answer and it’s 100% Worth It!
The Journey on Creating Successful Physics-Based Type Well Profiles
On August 14, 2003, LAX airport admitted entry for an eighteen-year-old boy who got very excited for an opportunity in the US. That boy was me. After spending more than half of my life in this great country, I feel like I have started something meaningful for the amazing O&G industry.
I have to say that graduating from Texas A&M and becoming a reservoir engineer working for two world-class operators were such blessing things an immigrant may have. Although gainfully employed, something felt off. I went to work each morning feeling perplexed. The one question weighing heavily on my mind was “What am I going to do with my type curves today”?
Occasionally, my boss stopped by and said something like “Duc, operator XYZ just increased their average lateral lengths, tightened up their spacing, and increased their proppant loading to 3000 lbs/ft. We need to update our type curves. John needs this evaluation done by end of the week.” If you are in the oil industry, you must smile.
My dilemma was not due to my inability to perform type curves analysis, but the analysis itself that the industry uses to construct type curves. In good conscience, I could not theoretically embrace the industry standard for forecasting. I asked myself “Did my education from Texas A&M and PE license make me qualified to do this kind of work rather than my non-engineering colleagues? They can do exactly what I do to the curves here and even better. To make the problem worse: If our company is spending a few billion dollars to buy in any of these assets and the future wells underperformed my type curves, am I going to be able to keep my job? Or are they going to audit my work and terminate my PE license due to lack of documentation and scientific backup for the results?”
After years of self-inflicted mental torment (I know you are wondering why I cared), I decided to go all-in and develop a type curves application based on science. Here was where the journey began, and the adage ignorance is bliss comes to mind. When I started USI, all I had in 2017 was a strong passion to resolve type curves issues. I started with the premise that if DCA (Decline Curve Analysis) could draw curves manually, a simulator can draw curves much better than people and reduce uncertainties by adding physical parameters associated with those curves. After a few years of doing simulation work, we decided to use CMOST to start “drawing” curves, and we thought we had the answer but after many setbacks, we decided that we were wrong. After working with more than ten companies including operators and reserves auditors, some who concluded (in the words of shark tank) “We’re out and come see us when you have a working product.” Other companies have been with us from day one.
In the early days, we partnered with a small operator in Midland Basin, and the reservoir engineer was a great classmate of mine at Texas A&M. He was a tremendous asset through the first trial phase of the product. Guess what? The trial failed miserably due to lack of technical knowledge and customer service. We matched the wells with our limited pre-run simulation library, but the match was only for a single well at constant bottom-hole pressures since we didn’t know how to control the models at various drawdowns using actual field histories.
One day, while in Midland trying to persuade a company to “believe” in our potential product (which had yet to prove itself viable) I got a phone call from my wife stating the lights were out. Oops and dang. I was so vested in this journey; that some aspects of my personal life were being overlooked.
Transitioning from engineer to entrepreneur, I had no idea how to follow up with a customer appropriately. The feedback from the first customers we had was turning into “we don’t see any value in renewing.” After losing our first customers, we found a few new clients who believed in our vision and were willing to take a chance on us. Once again, we slipped, stumbled, and failed. Our two biggest obstacles were technically related. One was our pre-run models were too simplified and couldn’t handle well-to-well interference. The other one was our ‘library’ of runs could never get big enough to cover all the scenarios for the wells that our clients had. What do you do when you have no working product, no immediate solutions and the money looks funny? How to survive through the pandemic and the downturn was still a miracle for us.
Getting better every day was our mantra. Through the last eight years of innovating and pushing through tough days, we finally came up with a new way that can benefit our industry at a large scale. The truth of the matter is: every existing reservoir workflow has its own flaws, whether due to lack of accuracy, or expensive, requiring intensive input data and time-consuming. Reservoir engineers always have to make a tough choice of which ones to give up on. There has never been a method where ‘better, faster, and low cost’ can coexist in one.
To solve complex challenges that we currently are dealing with, we must come up with complex solutions. Our approach shifts the fundamentals of production forecasting by generating curves based on pre-run simulations before applying curves directly on production data to establish trends. The method gives you critical insights on which combinations of physical parameters to use and minimize uncertainties with opposed to ‘completely guessing’ all the way through.
Selecting meaningful parameters to study was a tough and long process that we had to go through. We started with so many of them with low and high ranges in a single well model including: initial reservoir pressure, reservoir depth, bottom-hole flowing pressure, bubble point pressure, dew point pressure, pressure gradient, reservoir temperature, reservoir thickness, oil density, gas gravity, matrix and natural fracture permeability, non-fracture zone matrix permeability multiplier, vertical and horizontal permeability multipliers, rock matrix/natural fracture porosity, natural fracture spacing, matrix/hydraulic fracture initial water saturation, water-oil contact depth, matrix/natural fracture compressibility, lateral length, cluster spacing, well spacing, number of clusters, hydraulic fracture half-length/height/width/conductivity/permeability, hydraulic fracture compaction, relative permeability tables, and Pressure-Volume-Temperature (PVT) tables.
After several years of R&D work on the impact from each of these parameters, we revised and arranged them in a much more meaningful way into nine categories for our pre-run multi-well models. These subgroups of parameters are Reservoir, Completion, Pressure drawdown, “Neighboring well Influence” factors: Timing, Pressure, Completion, Spacing, Quantity of wells, and mostly importantly FDI factors.
Being a simulation engineer, I was never a big believer in machine learning and thought that it would never work. I was wrong. I assumed machine learning only can solve short term issues in production forecasting, I was wrong also. I would never imagine if we applied neural networks on pre-run simulated curves, how powerful the ‘combo’ models would be. The results and finding we got to date were beyond words. Combining the two really helps us resolve well interference and limited library runs mentioned above. More importantly, we now can ‘simulate and generate physics-based type curves in Spotfire within seconds’ compared to regular simulators which take days/weeks to make runs for large multi-well models.
Houston October 14, 2021
Duc Lam- Founder
To get more information about our product, please submit a form under Contact and we will get back with you as soon as possible.