【本周专题：机械采油】智能气举优化套路深 虚拟仪表给出绝杀

A Google search of “production optimization” returns about 2.2 million results. Clearly, it’s a well-discussed topic.

But, as some wag once asked about “diddy wah diddy,” what does it really mean?

Intellidynamics steps in with a useful definition: “Production optimization is essentially ‘production control,’ where you minimize, maximize or target production of oil, gas and perhaps water. For example, you can easily maximize or target the production of oil and/or gas while minimizing water, or run oil production and gas-oil ratio (GOR) to set points, to maintain reservoir energy. There are a myriad of alternative production objectives. Since each well, platform and field are quite different, a flexible means of controlling production is provided.”

Intelligent gas lift optimization. The company describes an application of this idea. Shell Malaysia, in conjunction with Petronas, gave this capability the name “iGLO” for Intelligent Gas Lift Optimization, when they applied it to the South Furious and Saint Joseph fields off the north shore of Kota Kinabalu, Sabah, Malaysia. In this project, [the company] modeled oil, gas and water production for each well on multiple platforms, and put those models online, in real time, as virtual meters. This gave them insights about where oil, gas and water were coming from in real time. The asset manager immediately discovered “lost oil.” Previous to this, test rates did not sum to production rates; a notable amount of oil had “gone missing.” But now, with visibility of production in real time, the asset manager could see how, and where, co-mingled streams’ interactions decreased overall production compared to test. A mystery was solved.

Perhaps “virtual meter” arched your eyebrows. According to the company, a virtual meter is a mathematical model that uses process conditions to estimate flowrates instead of using a physical meter. Virtual metering is a special case of a broader class: Virtual Sensors (space prevents elaboration here). In order to create and calibrate a virtual meter, you do, indeed, need flow but only for a while. Once calibrated, the virtual meter can then just use process conditions to make its estimates.

In the case of oil and gas production, physical meters are either on production separators measuring conjoined flow from many wells at the same time, or on test separators that measure flow for individual wells only occasionally and temporarily. During this “on-test” time, well operating conditions and production data are acquired for that particular well. These data are used to calibrate theoretical and/or data-driven, non-linear regression models. Once the cause-and-effect relations are captured, the model can be put online using the current well operating conditions. The model then outputs production estimates in real time. In this way, the asset team can see production from each well, even when not on test.

Why use virtual meters? The company has an answer for this question, as well: “Physical multi-phase meters that measure oil, gas and water are very expensive, more so if they are good for accounting purposes. They are also expensive to install. In the case of offshore platforms, they have to be ‘barged out’ or air-lifted to the platforms. Also, there may not be space to put a multiphase flowmeter into the tangled maze of pipes. Physical meters also need to be maintained, subject to corrosion, sand, plugging with wax, methyl hydrates, etc., which can be costly as well. If you wish to have such meters on each well, and considering there are typically five to 15 wells on each platform, the cost is prohibitive and causes the asset team to make ‘educated guesses’ about how much a well is producing. On the other hand, virtual meters are software, much more affordable, and much easier to create and deploy.”

Autonomously self-maintaining virtual meters. As the company notes, “…even virtual meters need to be maintained, because the wells change, and the reservoir depletes over time. In mathematics this is called a ‘non-stationary process,’ a process that changes over time.

“To maintain a virtual meter, we use the occasional well tests to compare the rate given by the model vs. the rate given by the physical meter, and make adjustments or re-field-fit the model. To do this we automatically monitor for such a well test to occur, capture the data, compare it to the estimates (oil, gas and water production), and either adapt the models to reduce their error or, if the error is large, remodel with the new data and perhaps some of the recent past tests. This all runs ‘autonomously’ (automatically, unattended). We also have the platform meters that measure the sum of all the operating wells. We can compare the sum of the virtual meter estimates to the sum of production meters from time to time and adapt the individual wells’ estimates using a variety of ‘sly’ algorithms. Sly? Yes, because you do not want to over-adapt to the platform meters, as it may cause a loss of ‘intuitiveness’ of the individual production estimates. This, too, is autonomous.”

By the way, the story of the company’s founding is surely relatable to many of you. “In 1979, founder Carl Cook was standing in front of a control panel of a manufacturing line, three stories tall and a city block long. The line was down for making reject product. He looked over a three-ring binder ‘expert system’ that told him that if he had Problem A, turn Knob #1 up, and if he had Problem B, turn Knob #1 down. The problem was he had Problems A and B, which suggested turning Knob #1 both up and down; not doing anything. He thought to himself, ‘There must be a way…’”