Case Study: Optimization Across a High Volume of Wells
Learning how to optimize across a high volume of wells is one of the key features of our Production Intelligence Solutions. In times of lower oil prices, reducing the production costs through optimization become more important. When we have a large volume of wells what do we do?
Solution: We synthesize physics-based simulation with machine learning-based solutions for optimization.
To explain this solution, we are going to tell the story of Mark, a gas lift expert working in the Woodford shale. Mark looks at the production rates of the well during the last 10 days. He wants to model the well and also optimize the gas injection rate on it. The well is shale-based, and hence there is a large variance of 60-90 barrels of oil production rate per day. There is also a variance in the Gas-Oil Ratios, Water cuts, and wellhead pressures. Mark is not entirely sure about the oil API gravity, Reservoir Pressure and the IPR of the well, although he has a rough estimate of that. He knows that gas is being injected at 0.4 MMSCFD and he generates all the possible simulations which match with the observed oil production rate and considers the variance in all the mentioned parameters (seen below).
In an attempt to save the gas, Mark wants to reduce the gas from 0.4 MMSCFD, to 0.2 MMSCFD. By looking at the graph above, this may look like a tedious and time-expansive process. To made an expert decision and developing the well model involves understanding all the underlying parameters, the well design, understanding the variance in the data, how to clean-up anomalies, and generate all the simulation. This may take a good part of the day for Mark, for one well! But using our simulation, Mark can expedite this process.
Let us take a look at what happens to the simulation when Mark reduces the gas injection rate from 0.4 to 0.2 MMSCFD. By making a change in gas injection rate, Mark has eliminated several simulation cases that were previously considered as possibilities. After elimination, we can see flat lines, and we can see curves. The flat lines tempt Mark to reduce gas injection further from 0.2 to 0 MMSCFD with no change in oil production rate.
Mark makes this change to 0 gas injection. We all may prefer 0 gas injection rate not just for gas savings but also for less maintenance. As we see over here, Mark realizes that after changing the gas injection rate to 0, there was a drop in oil production. This indicated that the well was operating on one of the curved lines. Taking into account, the economics of oil price and the gas cost, Mark decides to set the injection rate back up to 0.2 MMSCFD. He determines this to be the optimum gas injection rate at that point in time.
After the change, Mark has found that there are only 4 simulation possibilities remaining. Through the elimination of unlikely cases, Mark has found the economic optimum, and also has found out what is likely the underlying conditions for the current well. You too can be like Mark, and optimize across a high volume of wells.
If you want to learn more about using simulations to expedite the decisions for your digital oilfield, please watch our WebCast Replay entitled, “Top 3 Problems AI Solves in Artificial Lift Production” where Dr. Venkat Putcha, our Senior Data Scientist, and Tim Burke, VP of Operations and Client Services, discuss this decision case, among others, in detail. Click Here to access our webcast replays. If you would like to implement our Production Intelligence solutions, contact us to set up an appointment with our sales team to discuss easy implementation. Contact Us Today!