The future of data-driven analysis in exploration and production (E&P) will depend on whether it can add value in the field.
Four examples of what is possible were presented recently at the 2016 Unconventional Resources Technology Conference (URTEC) in San Antonio with the authors of papers posing questions such as:
- Does it matter if a lateral is drilled toe-up or toe-down?
- What are the changes in a fracturing design that will offer the biggest production payoff?
- Why has the drilling slowdown not depressed production from unconventional gas plays?
- What is the half-life of my field, and why should I care?
The stories below bring together advanced statistical analysis methods with multiple names: analytics, big data, machine learning, and even a “physio statistical engine for automatic stochastic production forecasting.”
All this new E&P math is aimed at identifying patterns and relationships that otherwise would be missed. But automatic and self-learning does not mean all-knowing.
For example, one of the programs used to analyze fracturing used facial recognition to classify pressure changes during each stage to sort them into different classes.
Production data revealed that one of the classes of fractures, characterized by a pressure bump near the end of the stage that could mean trouble, were often more productive than those that went exactly according to plan. But it was up to the completion team to figure out how to apply that observation.
The process is “brutally empirical,” said Roger Anderson, president of AKW Analytics, who led the study for Range Resources. He said the method is good at identifying what is happening, but it is unable to determine why or how it happened.
Toe-Up or Toe-Down: Does It Really Matter?
For years, an unresolved question for those drilling horizontal wells has been: Does it matter if it is toe-up or toe-down?
Horizontal wells generally follow an up or down slope following the most productive rock, which can mean large changes in elevation from the heel—the curve from vertical to horizontal—to the toe.
Results from modeling have not settled the matter. Recently, Devon Energy offered its response based on the performance of more than 300 similar wells drilled in the Cana-Woodford Shale in Oklahoma.
The comparison was possible because the company had mass-produced similar wells, giving it a large sample of wells with 4,800-ft laterals where it fractured 10 stages with 40 perforation clusters using 3.5 million lb of proppant located in a compact area with similar geology in the three depths studied.
The analysis was done in a way that ensured, as much as is statistically possible, that “only toe-up or toe-down could be affecting the production performance of the wells analyzed,” said Sam Browning, a reservoir engineering for Devon who delivered the paper at URTEC.
It concluded that toe-up wells produce more. And the greater the elevation change between the heel and the toe, the greater the impact.
“The more toe-up they are, they [wells] appear to be better, and the more toe-down is worse,” Browning said, adding “All the best wells were toe-up.”
Devon found that toe-down wells produced 25% less based on a year’s worth of production, he said. The difference was narrow in the early days of production and widened over time.
The results were compared based on the depth of the wellbore—shallow, middle, and deep—because of differences in the zones. For example, the condensate-rich production in the shallower zone is more likely to be affected by liquid holdup than the gas-prone deep zone.
The widest variation between toe-up and toe-down wells was seen in the middle zone where the elevation changes were greatest. In those toe-down wells, the elevation changes were nearly all in the 100–200 ft range, while the dips at other depths were less than 100 ft.
The paper predicted that an extreme toe-down well in the lowest-pressure area would produce 30% less over its life.
The paper suggested toe-up wells may perform better by offering a gravity boost for the liquids-rich flow, and toe-down can allow liquid to build up in the production zone, hindering the flow.
Browning is planning to come back to these wells to see how they are doing in the future, which will address questions about whether changes that come with age, such as increased water production requiring pumping, will alter the conclusion.
Devon is using this study when planning new wells in other areas with similar conditions, such as projects to drill infill wells in spots with relatively low pressure to overcome liquid buildup. The location of the well, though, may require drilling a well toe-down to stay within the most productive rock.
And the sample size in this case is too small to generalize about all horizontal wells. But based on the Devon study, the potential reward for knowing if toe-up is better or worse can be big enough to justify the cost of a study to answer the question.
URTEC 2461175 Effects of Toe-Up versus Toe-Down Wellbore Trajectories on Production Performance in the Cana Woodford by S. Browning and R. Jayakumar, Devon Energy.
Is the Struggle Really Worth It?
When it comes to pumping a proppant into a fracture, a job that goes according to plan may not be the best option.
A statistical analysis of fracturing using machine learning found that the smoothest fracturing stages often underperformed those where the operators struggled to get all the sand into the formation.
The study presented by AKW Analytics at URTEC used a system comprising multiple software programs to seek out telling patterns in several streams of data from 1,800 fracture stages in 156 wells completed by Range Resources in the Marcellus Shale.
The study focused on two groups of classes of fractures: ones where the pressure rose before it completed pumping the planned amount of proppant, raising concerns that an excessive amount of sand might block the fracture by causing a screenout, and another class where the pressure hardly changed because “everything went as scheduled,” according to the paper.
But the production from the fractures that went smoothly was often not as good as those that required a struggle. “Perfect means you will not do as well if you do not struggle to get in that last sand,” said Roger Anderson, president of AKW Analytics, who delivered the paper on a 4-year study for Range. “Perfect may mean the sand (volume) was short.”
The difference among fracture classes, which was only noticed in 69 of the wells, was one of many variables studied to see which factors have the biggest impact on production.
The goal was to learn how to predict future performance when completion teams are considering which variables will have the most effect during the design stage.
The multivariable analysis showed an oil field is no place for a statistician looking for clean, consistent data sets over time. Prior to the last year of the study, Range Resources revamped its completion method. The pounds of sand placed per foot nearly tripled, with coarse sand used and more stages in longer laterals drilled deeper in the formation.
The paper points out that “determining which (change) led to the greatest improvement is difficult because” there were so many big ones. The bottom line, though, was positive, with the new approach yielding a “remarkable increase in condensate and gas production per well in 2013,” the paper said. And Anderson pointed out the changes were generally in line with the study’s conclusions about which factors matter most.
The most influential ones were not surprising—the percentage of the wellbore in the targeted reservoir rock, the total volume of proppant placed, and the number of stages were among them—but it also quantified the potential impact and predicted how different combinations of attributes will perform.
By using 34 attributes, for example, Anderson said the model was 78% accurate, with a wide error range. With more data to consider and time to learn, the system could improve its accuracy. “It is not something you would want to report to the SEC; it is not good enough yet,” he said, adding, “but it is better than guessing.”
URTEC 2430481 Using Machine Learning to Identify the Highest Wet Gas Producing Mix of Hydraulic Fracturing Classes and Technology Improvements in the Marcellus Shale by R. Anderson, B. Xie, L. Wu, AKW Analytics et al.
URTEC 2426612 Petroleum Analytics Learning Machine To Forecast Production in the West Gas Marcellus Shale by R. Anderson, B. Xie, L. Wu, AKW Analytics et al.
Why Does Drilling Drop, Not Kill, Gas Production?
Shale wells are known for their rapid production declines. So when drilling activity dropped in 2008 and again in 2014, steep declines were widely predicted.
But that is not how it has played out in the three biggest US shale gas plays—the Barnett, Haynesville, and Marcellus—where production has held up even though only a few rigs have been running.
The reason why the expectation and the realization have been different is due to the industry’s increasing productivity and the fact that thousands of unconventional wells perform differently than a single well, according to Philippe Charlez, a senior technical advisor to Total, who delivered a paper on his research on resilience at URTEC.
Total’s Unconventional Factory Development Simulator was used to study the decline rate of a large shale play with 3,000 wells producing 1 Bcf/D of gas. The results showed that the impact of the rapid early decline rate of new wells can be muted by the slow, steady production from thousands of older wells drilled over many years.
While the paper said it takes a fleet of drilling rigs to reach that peak—16 rigs working for years—once at that peak level, only four are required to maintain that level for many years. In practice, operators are likely to add rigs when gas prices are high enough to make it an attractive investment. “When the price is relatively high, I drill a lot to feed my well portfolio. When prices are low, I live on existing wells,” Charlez said.
There is a significant reward for productivity gains increasing the estimated ultimate recovery (EUR). “If it is higher, I need less rigs to maintain the plateau rate,” he said. And if an operator increases the EUR by about 20% over a 3-year period, a simulation concluded the increased productivity would be possible to maintain the production plateau for 10 years with only two rigs.
The simulator was also used to predict what would happen in the three biggest US gas shale plays if drilling stopped. It found that the Barnett was the most resilient based on the expected rate of decline, the Haynesville the least, and the Marcellus in between.
“The Barnett is very resilient. There is a large well portfolio,” drilled over many years, Charlez said.
If drilling stopped for 10 years, the Barnett production would be down 50% while the Haynesville would have dropped to zero by 2025, according to the study that said Marcellus would drop nearly 75% if there was no drilling during that period.
As the oldest shale play, the Barnett has the edge because about 15,000 wells have been drilled there vs. 3,700 in the Haynesville, and the Barnett has more, older wells whose steady output reduces the impact of the sharp declines seen in new wells.
The net effect of a mix of prolific fast-declining young wells and low-producing steady old wells is roughly analogous to a diversified investment portfolio with volatile high-performing stocks balanced by the modest, steady cash flow from bonds.
URTEC 2439429 Resilience of the US Shale Production to the Collapse of Oil & Gas Prices by P. Charlez, Total, and P. Delfiner, PetroDecisions.
What is the Half-Life of That Shale Play?
Half-life is a commonly used way to measure the life of radioactive materials, but not oil fields.
The founder of a data-driven forecasting company, though, argues that it is a useful benchmark for companies with substantial production from unconventional oil fields, because it can be used to estimate when a company will need to spend money to sustain production at a level that satisfies investor expectations.
To make her case, Heidi Kuzma, founder and chief technical officer of BetaZi, applied its statistical analysis method to forecast production from more than 13,000 wells in North Dakota and concluded it would take 3 years and 4 months for the output of all those wells to decline 50%.
“Somehow, if production is to be brought up to its peak 2015 levels, half of that production must be replaced by the end of 2018. Worst case scenario, that means another 6,000 or 7,000 wells,” need to be drilled and completed, said Kuzma, a geophysicist by training. “That is a huge amount of capital.”
The actual statewide decline rate will not be that fast, because while the calculation assumed no drilling, there were 27 drilling rigs working in North Dakota in late August, according to the Baker Hughes rig count. Those wells are for companies focused on speeding the construction of significantly more productive wells.
For individual operators, however, the half-life of their assets is shorter. The pressure is greatest for those who drilled most of their wells at the tail end of the boom, which ended during the second half of 2014.
“There are already a significant number of operators whose Bakken production is very close to half of what it was at its peak in 2015,” she said. Many of those cannot afford to drill and complete wells. “In this case, depletion might be a function not of the asset, but of the capital needed to exploit it,” Kuzma said.
BetaZi’s statistical approach estimated the decline rate by inputting monthly oil, gas, and water production figures reported to the state of North Dakota for each well, and also the number of days the well was flowing. The company’s predictions were created using its proprietary “physio statistical engine for automatic stochastic production forecasting.”
Estimates of reserves and ultimate recoveries do not figure into the calculation. In shale, it appears the amount of oil produced is a small percentage of what is in the ground. Actual recoveries will depend on the future price of oil, and whether improved techniques are developed and applied.
“Reserves and EUR [estimated ultimate recovery] are estimated numbers that are difficult to test. They constantly change,” Kuzma said.
Production decline rates have always been tracked, but are not featured in financial reporting. Predicting long-term decline rates from unconventional wells is a challenge. The proper method for predicting the long-term decline rate for unconventional wells is not settled among petroleum engineers. And the skills and methods of those drilling and completing wells varies widely among operators, which also have varying approaches, and budgets, for maintaining production.
Based on the conclusions of Kuzma’s paper, and one presented by Phillippe Charlez, a senior technical advisor for Total, long-term production predictions are tricky.
Both estimated production on a well by well basis and combined it using their company’s proprietary software engine. He drew his information from decline curve analysis, while she statistically predicted future monthly production rates. And both said their predictions were in line with actual production results.
Charlez concluded that if “no new development activity occurred, it would take 10 years for Bakken production to drop 50%,” which is three times longer than Kuzma’s prediction. His estimate was mentioned at the end of a study focused on major gas plays, where he said the longer histories allow more dependable statistical analysis.
Given the many variables, Kuzma said it is hard to prove who is right. “For the moment, any purely data-driven forecast that claims to be accurate beyond 6 or 7 years for unconventional production is a bit suspect, since we just don’t have that much experience,” she said, adding, “let’s get back together in 10 years.”
URTEC 2461672 Blinking Out: North Dakota Without Capital for Replacing Production by H. Kuzma, C. Eklund, T. Rapp et al., BetaZi.
