Optimizing big data in the oil field

Long-term operating programs in the oil field require many decisions. Selection of timing, approach alternatives, and choices of scale – just to name a few – collectively present a daunting matrix of decision types to a program over its lifetime. Some programs are built on a cascading sequence of decisions for a complex and changing asset base, and any single decision in the chain can result in poor performance.

As operators move into more challenging environments such as Africa and Asia, experience alone is insufficient to ensure optimal operations. A number of new technologies are available to assist operators in making their most complex decisions. Notable among them is “big data” – the ability to apply computation across very large datasets quickly and often in real time. The energy industry is on the cusp of developing applications using big data tailored for upstream operations to make logical predictions and solve complex problems. These applications will reduce the need for operators to rely solely on intuition and clunky spreadsheet-based solutions to solve convoluted problems.

FIGURE 1. The well intervention cycle starts with the choice of one intervention among many opportunities - from a 'hopper' of candidates to execution to outcome. (Images courtesy of Business Laboratory LLC)

When experience is combined with methods specifically designed to cut through the intricacy of the decision-making process, the right solution may be found. Optimization is one such method.

There are numerous opportunities to apply optimization in the oil and gas industry, ranging from operating theater logistics to the efficient distribution of oilfield materials.

What is optimization?

The term “optimization” is used frequently with multiple overlapping definitions and often as a synonym for considering all available options. In this article it is used to describe the practice of mathematical optimization, a process of structuring a problem into a format that allows a computer model to explore all possible solutions so it can find the one that best achieves the desired outcome.

Any optimization model consists of three fundamental features:

• The objective function, representing the desired outcome;
• Constraints limiting the solution to those that are feasible; and
• Variables, those values that the model can change to affect the outcome.

In energy operations, operators usually set their objectives to increase production, grow reserves, or free up cash flow. Companies use resources that have constraints, primarily physical capacity limitations. Finally, the areas of control, such as the number of assets or the rate of work, represent decision points in the system. Creating a mathematical optimization of an energy problem involves mapping these real-world values to their model counterparts. Engineers choose wells for intervention based on a variety of factors – the age of the well, ease of access, adjacent crews, supervision and engineering, and estimated production lift. Often, these factors are worked out on a spreadsheet, but in reality there are far too many considerations for which to account when choosing the optimal sequence that will result in the greatest overall production increase. Companies benefit when using a well-designed optimization process that will enable them to control more factors than if they were to use a traditional analysis.

Preparing for optimization

FIGURE 2. This graph shows the incremental production realized over a fixed period of time from a well intervention operation.

In the following example, one major oil and gas company successfully applied mathematical optimization to maximize production from its intervention operations. It sought to improve the performance of its well intervention team, and it wanted an optimization model to leverage the experience of the field team while reducing the manual workload to construct the optimal sequence.

With a portfolio of 16,000 oil and gas wells across North America and thousands of opportunities for intervention at any given point in time, the company had to determine the right order of well interventions given its rig inventory, geography, crew availability, specific characteristics of the individual wells, and the extent of the intervention required. The difference in value between an optimized sequence and a random sequence was many millions of dollars annually, according to one estimate.

Before it could begin the optimization process, the company needed to translate the well intervention problem into the optimization model’s three fundamental features:

Objective. Determine the best way to maximize the production from interventions. The current baseline of performance was poor. After simulating intervention operations over a fixed period of time given the number of rigs and supervision available, the company observed the unique characteristics of each well and noted how each required a different form of intervention. The result of the simulation was an incremental production curve as shown in Figure 2.

Constraints. Consider the number of rigs, crews, and operator resources available to work on intervention jobs.

Variables. Keep in mind that the model may choose any sequence among the billions of possible sequences across the well portfolio. The variables reviewed in this case study represent the “single file” choice of well interventions over time, organized by rig.

How optimization works

Once the problem was sufficiently structured to create a model, the optimization “engine” considered the combinations of factors that could provide potential solutions to finding the optimal result. Modern methods coupled with today’s computing power often render solutions to even the most complex problems in a matter of seconds or minutes. In this case, the result was a sequence of well interventions as shown in Figure 3.

Comprehensive simulation can be difficult to design and implement in practice and requires specific experience to devise the objective function, consider the constraints, and interpret the outputs of the optimization model. If done right, it can substantially contribute to the operational and financial success of the field.

FIGURE 3. This table shows the result of an optimization run with the wells for intervention organized by available rig.

The benefit of an optimization model is not just the mathematical solution itself. The inherent debate and critical thinking that accompany the process of structuring the problem generate insights among the team that can be as valuable as the model itself.

Optimization can be applied to any problem in oilfield operations that has similar attributes: complexity, scale, a defined objective, constrained resources, and large datasets. Field operations are constantly changing, which makes the ability to reoptimize quickly and frequently vital to any optimization model. Given the availability of technologies in use today, optimization is within reach of small and mid-sized firms, not just the industry giants.

Solve tomorrow’s challenges today

As recently as a few years ago, a solution that optimizes operations across several thousand assets – each with a substantial profile of associated data – and in close to real time would have been unfeasible, forcing companies to compromise on the accuracy and quality of the solution.

The big data technologies available today allow us to overcome these limitations, giving us solutions with far more accuracy than before. With the complexity and scale of oilfield operations constantly increasing, the industry should embrace big data as an important element in the problem-solving tool bag.