By Atanu Basu, Daniel Mohan and Marc Marshall, Ayata

Unless you have been living under a rock – no pun intended – you have experienced the buzz in today’s business media about the Internet of Things and Big Data Analytics. Some of you may have even wondered how would these macro trends impact finding and extracting more oil faster, cheaper and safer. This article explores how a technology that sits in the middle of these two trends is poised to transform oil and gas operations, with a focus on unconventional oil development.

Industrial Internet of Things

The CXO’s of GE, Cisco, Siemens and like are onto something when they frequently tell the press that ”tens of trillions” of dollars will be added to the global GDP over the coming decade through the Internet of Things. To simplify, the Internet of Things – more accurately called the Internet of Everything by Cisco – means there will be sensors everywhere and in everything (and on everyone) and these sensors will connect and communicate over networks facilitating decisions and actions that haven’t been possible before. The Industrial Internet of Things, a sub-segment of the Internet of Things, is focused on the impact of this phenomenon on different industries, including energy.

In oil and gas, the amount of data that we are collecting today is much more in quantity and diversity than what we have done in the past. And the situation will only get better – or worse, depending on who you ask – over time as we will continue to generate more data (Volume) of different types (Variety) at a faster pace (Velocity) from every well.

Unconventional energy development requires more than 10x (might be over 100x) the number of wells and more than 30x the number of completions than its conventional counterparts. Subtle changes in well placement, spacing and completions techniques often determine economic viability. With the shortage of unconventional engineering talent and many more wells and complex completions, operators must find a way to better analyze dramatically increasing amounts – and types – of data.

Three Shale Realities

The oil and gas industry has been a Cinderella story for the past several years for the U.S. Horizontal drilling and hydraulic fracturing have made it possible to profitably produce oil and gas from unconventional sources, primarily shale rocks. The industry has been continuously learning, adapting and improving exploration and production from the shale plays. Three themes seem to have surfaced.

Rapid Decline: Shale wells decline much faster than conventional wells (charts below). To keep the U.S. energy boom going, the oil and gas industry has to keep putting new holes in the ground, which obviously has side effects with respect to capital, environment, citizens, policy makers and more.

Productivity and Cost: Despite all the amazing accomplishments of horizontal drilling and hydraulic fracturing in economically producing oil and gas from shale rocks, hydraulic fracturing is still an inefficient process. The EUR in shale plays has been 3% to 5% for most operators. Moreover, up to 30% variation in well performance has been reported among companies operating in similar subsurface conditions. Emerging technologies such as fiber optic sensing combined with existing diagnostic tools such as microseismic are now showing us that the current fracturing process isn't optimized. Many times only a small fraction of the reservoir is stimulated with each fracture treatment.

While we keep repeating – often robotically – that it is all about the rock, the truth is it is also about knowing what to do to the rock ‘and’ verifying if what we are doing to the rock is working as we intended it to. The last two have turned out to be much easier said than done.

Environmental Concerns: A lot has already been written and said about the possible environmental concerns – earthquake, water contamination, etc. – surrounding hydraulic fracturing. A quick Google search – fracking environment – generates more than 7 million results! Reducing environmental footprint and improving health and safety are among the top agenda items of executives at operators and energy services companies.

Shale – Mindboggling Data Complexity

It is really difficult to make hydraulic fracturing more effective and safer at the same time. There is a lot going on – and the engineering keeps improving. Enormous amounts of data are used and generated during exploration and production of unconventional energy. In addition, these datasets are usually in different formats – some examples follow:

--Images: Seismic (1-D, 2-D, 3-D, Time Lapse 3-D), Microseismic, Well Logs, Mud Logs, Offset Logs

--Sounds: Of drilling, fracking, completion and production recorded by fiber optic sensors (distributed acoustic sensing, [DAS])

--Videos: Cameras monitoring downhole fluid flow, videos of Pressure-Temperature-Strain data recorded by fiber optic sensors

--Texts: Completion procedures, Core data, notes taken by drillers, pumpers and engineers

--Numbers: Production data, Artificial Lift data, Fracture Stage data

How does one take these disparate datasets and make sense of them, together? To improve EUR and reduce environmental footprint, extracting actionable information after combining all these datasets is crucial. As our industry gets more into ‘shale manufacturing’ and as exploration and production (E&P) costs rise, an operator’s ability to improve production performance while becoming safer is instrumental to its success in the highly competitive marketplace. The ability to make accurate investment/drilling/fracturing/completion decisions is a must-have core competency. It is not always about producing the most oil, but producing the most profitable well with the highest rate of return (ROR).

Could Industrial Internet of Things and Big Data Analytics help with these challenges?

Artificial Lifts – Uncertainty Is The Norm

Electric submersible pumps (ESPs) and other artificial lift mechanisms play a pivotal role in oil and gas production as they pull hydrocarbons from underground to the surface. According to GE, there are more than 130,000 ESPs installed worldwide today, and these ESPs account for more than 60% of global oil production. The Achilles’ heel of ESPs is that they fail unpredictably causing unexpected production loss – which sometimes result in missed forecasts to the market.

As the artificial lift vendors embed different types of sensors in their equipment to measure anything and everything, how would we use the data generated from these sensors and other sources to predict and preempt production misses?

Google-like Technologies In Oil And Gas

In the world (mostly) outside oil and gas, software is ushering in an era unlike anything we have ever seen. Google Car – autonomous vehicle – is safer than the ones you and I drive. Computers today can easily beat human champions in the "Jeopardy!" game show and predict Supreme Court decisions more accurately than legal scholars do.

A self-driving car sees, hears, understands, decides and acts – and then repeats these tasks as what’s in front of the car, what’s around the car and what’s behind the car change. It is a marvel of hardware and software, working in synergy to accomplish a feat that someday will change transportation – and related industries – as we know them. Facebook algorithms will soon be able to recognize your pictures – even from those days when you had the funny haircut – more accurately than your loved ones can. In another recent study, computers were able to accurately discern who is faking pain and who isn’t 85% of the time – human pickers were barely better than a coin toss! Machines today can accurately identify images, including photos of a particular breed of dog.

Image: Kang Lee, Marian Bartlett. The woman on the left is faking pain, the other two are not.

As an operator, have you considered using computer vision, machine learning, natural language processing, operations research, signal processing and other technologies to interpret all the data (images, videos, texts, sounds and numbers) you have?

Machines Prescribing Recipes

Today, most operators have drilled only a fraction of their unconventional acreage. They learned valuable lessons – and collected invaluable data – from these drilled/fractured/completed/producing wells. Now the question becomes how can an operator extract and use the insights lurking in all these datasets to improve production – while reducing cost – from its remaining acreage positions. Enter prescriptive analytics. The premise is to make every well the best it can be by using every dataset at hand at the time of drilling this well.

Over 80% of world’s data today is unstructured – videos, images, sounds and texts. Making critical decisions based on only numbers (structured data) in today’s complex business and operational environments will often lead to disappointments or worse. Prescriptive Analytics takes into account all data to paint a more complete and more coherent picture of the future – and then makes this future actionable. It answers what will happen, when, why and how to take advantage of this predicted future without disrupting other priorities.

As more wells come online and new – and more – datasets become available, Prescriptive Analytics software adapts automatically so the predictions and the prescriptions for new wells stay relevant and accurate.

Prescriptive Analytics can offer data-driven answers to mission critical questions – some listed below – so the shale operators can compress the learning curve and make better decisions, faster. Less trial and error will lead to higher ROR.

--Which variables have the greatest impact on production? Why? Which of these variables do we control?

--How much production is attributed to the reservoir, and how much is due to controllable variables?

--What is the optimal configuration of controllable variables given a specific set of reservoir characteristics?

--How closely should we space wells? Why? Do we have stage overlap? Formation containment?

--Does the order in which we treat and/or produce adjacent wells matter? Do we have interference?

--Which stages and clusters were treated effectively? Treated as expected? Why?

--Which stages are producing? Producing as expected? Which are not? Why?

--What instrumentation should be incorporated into my completion design? Why?

--How should a well be produced to maximize its lifetime value? Why?

--How should we modify our completion designs and production models to more accurately reflect reality?

For artificial lifts, prescriptive maintenance can offer operators more accurate visibility into future failures and prescribe appropriate preemptive (or preventive) measures. Prescriptive analytics here must take into account data from the system and the surroundings, together: the pump, the fluid composition going through the pump, and the subsurface in which the pump is operating.

Machines AND People

We are not talking about replacing people with machines here. It is about complementing people with machines so each constituency can contribute using its core competencies. Machines can continuously process vast amounts of disparate batch and streaming data, discover patterns, provide insights, predictions and prescriptions that are always up-to-date. Machines enable engineers to better understand and optimize initial well design decisions, improve standard practices and use the machine as their 24/7, data-driven guide in continuously optimizing production.

The good news is the pioneers in the energy business are starting to pay attention to prescriptive analytics. Google-like technologies are finally arriving to improve some of the most expensive, in more ways than one, decisions in the planet.

Atanu Basu is the CEO of Ayata, a software company that invented and refined prescriptive analytics over 10 years. Daniel Mohan is the senior vice president of sales and marketing and Marc Marshall is the senior vice president of engineering at Ayata.