Shale operators are searching for insights to get the most value from hydraulic fracturing at a time of low oil prices.
Discovering these insights requires a closer analysis of well data in order to make more cost-effective field development decisions. “We are often working with short-term data and need to find a way to turn that into long-term well performance,” said Craig Cipolla, a senior completions engineering adviser at Hess, who spoke at the recent 2015 SPE Hydraulic Fracturing Technology Conference in The Woodlands, Texas.
Companies such as Hess and Noble Energy have invested in drilling test wells with hopes of recuperating costs by optimizing future treatments. “These [studies] require significant resources and manpower, and we can only do a limited number of these, if any at all in some cases,” Cipolla said.
Southwestern Energy has taken a different approach. It takes less expensive measurements and feeds them into models that are custom-built to be more predictive for the given play. The company already has made cost-saving adjustments to its operations based on the results of its models.
Each approach aims to make fracturing more profitable, whether by improving recovery or finding ways to achieve the same level of production at a lower cost.
Productivity Index
Southwestern has created long-term flow simulations of its Marcellus Shale gas wells from what reservoir engineering manager and technical lead Michael Lynch called “nearly free data.”
His team learned from simulations that proppant loading was one of the most important variables for productivity. Wells with high proppant loading overperformed by 40% compared with the company’s 30-well production average in the Marcellus, and those with low proppant loading underperformed by 20%.
The team also found that close spacing between fracture stages in its Marcellus assets did not significantly drive productivity. So to save money, the company has been using a wider spacing of up to 500 ft and more proppant per stage.
The main measurement that Lynch and his colleagues use to populate flow simulations is the initial productivity index, which is determined by shutting the well in and slowly opening the choke over the course of 2 to 3 days after breaking through to gas.
Lynch said such early production data may be used to make accurate predictions about the future performance of unconventional gas wells. “The single best way to learn about your completion is to have numerous shut-ins and rate changes, then mine that data, put it into a flow simulator, and I think you will learn more about your completion efficiency than anything else out there,” he said.
The data are processed using a simulator built specifically for the asset. “We had to tear it down and rebuild it, over and over again, before we would get the calibrated match we wanted,” Lynch said.
The predicted performance can be cross-plotted with other well variables to learn what sets productive wells apart from less productive ones, aiding in more economical fracture designs.
Intensive Data Acquisition
Hess hopes that the test wells it has been drilling at its Bakken Shale acreage will lead to better decisions about the number of fracture stages and volume of proppant used. However, drawing inferences about what improves production based on these wells can be complicated.
The company began its studies by investigating simple, 10-stage fracturing jobs using ball drop systems and uncemented liners. Cipolla said that the 90-day production from wells in the area using this method averages 35,000 bbl, but that of individual wells ranges from 10,000 bbl to more than 40,000 bbl.
This large spread means that it may take more than a dozen wells to confirm that one treatment offers a significant improvement over another. The number of wells required climbs when experiments are performed to correlate a change in stimulation with a modest increase in production. “If we’re trying to identify reliably a 10% difference [in production], we may need up to 60 wells,” said Cipolla.
The decisions made as a result of the new data may increase production and profitability, but initial costs can be high. “Our models can be predictive, but they need a lot of input,” he said. While drilling dozens of wells has been affordable for Hess, which has been in the Bakken for more than a decade, the cost is prohibitive for smaller operators.
With each increase in the number of stages, Hess’ tests showed that production improved as the count reached into the 30s. But at a certain point, the average production per stage dropped, Cipolla said.
During tests, an increase from 10 to 38 stages dramatically raised the 90-day production (by approximately 2.4 times), yet the increment was a nonlinear one, something Hess could not easily explain. This complexity is compounded by the fact that multiple variables often change when stage count is increased, such as the amount of proppant per stage.
When test results are not clear, the next step is to flesh out the data set with more wells, which Hess is in the process of doing, Cipolla said.
The company is also experimenting with other tests that use fewer wells and involve the collection of microseismic data.
The Underground Lab
Noble Energy’s fracturing optimization program benefits from offset drilling as well, but it looks for an edge in the quality of data collected, as opposed to the quantity of wells.
“The question we are answering is ‘Is it possible to get more concrete and reliable information from a single pilot [well], or one or two wells, that can be used to make decisions?’” said Dave Koskella, exploration and reservoir systems manager at Noble.
Using a larger array of sensors and gauges than Hess, Noble seeks to answer to the why behind changes in downhole variables by providing a more complete picture with complexes of research wells called underground laboratories.
In the first underground lab, sensors were used to collect information on how the environment reacted to fracturing in general, rather than to any type of treatment.
Noble’s second lab, which is located in the Niobrara Shale, in Colorado, aims to investigate how different stimulation methods compare with one another in a similar environment. Two horizontal wells were drilled as producers, in addition to a number of nonproducing satellite wells, which are strictly used for data collection.
Main measurements collected from the heavily monitored producing wells include distributed borehole temperature, toe and heel pressure, microseismic data, and uninterrupted production logging with fiber-optic line.
In the lab’s main experiment, Noble compared its preferred hybrid stimulation using 200-ft spacing with a hybrid method using 240-ft spacing, a hybrid with more sand at 240 ft, a hybrid with intermediate ceramic proppant at 240 ft, and lastly, slickwater.
The methods were used three times per lateral, except for the original hybrid, which was used twice per lateral. To control for geological variables, each of the two wells used methods in opposite orders.
Once the wells were fractured and placed on production, logging showed that Noble’s typical hybrid method resulted in the most consistent production. Slickwater performed the worst on a barrel-per-foot-per-day basis; yet some slickwater perforations outperformed the hybrid method.
Noting that slickwater was cheaper than the standard hybrid design, Koskella said that such test results are only part of fracturing optimization, and that the effectiveness of each stimulation treatment must be weighed against its cost to determine the choice that will achieve the most economical outcome overall. Slickwater fracturing, for example, requires a larger volume of water than gelled fracturing treatments, a factor that entails trucking in water in addition to the upfront cost of chemicals.