IT Comes of Age in Oil and Gas

“There are more things in heaven and earth, Horatio, than are dreamt of in your philosophy.” William Shakespeare: Hamlet
E&P companies are adopting more sensor-based technologies than ever before, largely to address the challenges inherent in exploiting remote and unconventional resources.  As a result, corresponding IT infrastructures are evolving to enable an increase in availability of raw data from disparate engineering silos which, in turn, feed analytical workflows and real-time decision making.

The concept of digital oilfields, built on this sensor-based technology, has matured rapidly over the past decade. Cloud services, satellite communications, powerful analytic software and soft computing techniques are just some of the important components that can result in increased operational visibility, reduced uncertainty and productivity maximization.

In practice, a digital oilfield with intelligent wells can easily overwhelm today’s information technology (IT) infrastructure because of high data volume and velocity.  A veritable tsunami of disparate data defines a complex, heterogeneous landscape such as a reservoir-well-facility integrated system. These high dimensionality data are supplemented by unstructured data originating from field notes and social media activity.  Not only are mobile devices proving to be valuable in field operations, but, with the advances in cloud computing and network performance, mobile is becoming the preferred method for accessing E&P data.  Due to these operational and market innovations, we are ideally positioned to marry soft computing methodologies to the traditional deterministic and interpretive approaches.

Recent developments in computational intelligence, in the area of machine learning in particular, have greatly expanded the capabilities of empirical modeling. The discipline which encompasses these new approaches is called data driven modeling (DDM) and it is based on analyzing the data within a system. One of the focal points inherent in DDM is to discover connections between the system state variables (input and output) without explicit knowledge of the physical behavior of the system. The wave of technologies washing up across the E&P value chain predicates the role of data driven methodologies focused on computational intelligence (CI) and machine learning (ML). The former embraces the family of neural networks, fuzzy systems and evolutionary computing. Data driven modeling is therefore focused on CI and ML methods that supplement or replace numerical models based on first principals. (See Figure 1)


Figure 1: Data Driven Model Supplementing existing processes

At the core of data driven modeling is a process known as SEMMA. The acronym – Sample, Explore, Modify, Model, Assess – describes a comprehensive process of conducting data mining. Beginning with a statistically representative sample of your data, the SEMMA approach enables a user to:

  • explore statistical and visualization techniques
  • select and transform the most significant predictive variables
  • model the variables to predict outcomes
  • confirm a model’s accuracy

This process is thorough and robust, relying heavily on mature data management solutions. As an example, developing an effective model for predicting and mitigating against drilling events is largely dependent on the availability and quality of data. So what are some of the barriers to creating an operational solution based on available data?

  1. Silos: Simply put, not all of the data are stored in the same location. Accessing all of the necessary data in an efficient manner that enables implementation of real-time solutions can be difficult to achieve based on existing and rigid IT architectures. One of the most effective ways to overcome this challenge is to take a data federation route for building and deploying analytical solutions over a data integration solution. Managing data by reference allows analytical solutions to connect multiple sources toward a single operational need without the need to employ expensive and invasive hardware and software.
  2. Quality: One of the most common statements around analytics is that the data are of poor quality and cannot be trusted. “Garbage in, Garbage out.” There are a multitude of automated data quality solutions that can be implemented toward managing data quality, but the levels of quality required can often be misunderstood. Data driven models do not always require data to be an absolute value of what is being measured. As an example of this let us consider the hookload sensor. The measurement from this sensor is considered at best a guide of how much weight is suspended from the hoisting system, and many people dismiss its value in deriving insight. The important difference in a data driven model, is that instead of considering the sensors absolute value, we consider it as an observation within the context of multiple inputs and outputs. The sensor may not provide an accurate measurement of a singularity, but it can reflect changes in the relationships between other measurements found inside this solution. As such the most important quality function in this case is its relative value within the context of the model.
  3. Inconsistency: Nobody knows which data are best – There is no “golden copy” or “golden record”. Again the federated model of data by reference can help us overcome this challenge. By agreeing on a single source in the federated model we don’t have to implement a time consuming data governance solution. All that is required is a decision on which reference to use and a process to be followed.

With the growth of while drilling and downhole distributed sensor systems, we now have a wealth of data available to us that not only calibrate deterministic models but also feed data driven models. These soft computing models constrained by first principles can surface hidden patterns in a complex, multivariate subsurface system (see Figure 2), aggregating both static and dynamic data to optimize reservoir management.


Figure 2: Complex Subsurface relationships

To attain the most efficient completion strategy in the unconventional reservoirs, it is necessary to perform a multivariate analytical suite of workflows that identify the most important parameters to impact performance. Neural networks were developed to optimize fracture design and identify geologic sweet spots, built on the enterprise data assets across the unconventional fields. The data driven methodologies form the arterial conduits of a digitized IT architecture that direct data in real-time through models developed on the rich enterprise historical data to identify operational parameters that maximize performance and optimize well integrity. The data driven models are constrained by engineering concepts to enhance both design and operational efficiency of a completion strategy. The hybrid approach provides solutions to operational questions to formulate completions optimized in diverse unconventional plays based on both engineering experience and the rich array of data. The results generated by the analytical workflows are fed back into the enterprise data assets, supplementing and cultivating a robust suite of data on which actionable knowledge is ascertained in a digitized environment. The operationalized model is encapsulated in the IT architecture as one node utilized by extant software solutions (see Figure 3).


Figure 3: Advanced Neural models encapsulated into operational solutions

With an eye on the future, we can now consider how advanced analytics can be deployed at the asset level while maintaining the leverage that volumes of data at the enterprise level promise. The challenge is in deploying low frequency high latency modeling solutions into a real-time asset that requires high frequency and low latency decision support. This is where event stream processing solutions can bridge the technology gap. By enabling deployment of complex models that are infrequently updated into an intelligent high-speed modeling environment, we can now deliver asset level decision support based on enterprise level analytics (See Figure 4).


Figure 4: Enterprise analytics applied in real-time at the asset.


IT has come of age in oil and gas, with innovative E&P companies increasing operational visibility, reducing uncertainty and maximizing productivity by enabling real-time decision making through analytics, which form the basis for competition in the 21st century.