We recently benchmarked our Actenum Rig Activity Scheduling (RAS) application against a manually-derived operational rig schedule obtained from one of our clients, a large oil producer. We achieved a 6% reduction in rig transportation costs and a 38% reduction in time to production, which translated into more than a 15% increase in net present value.

Impressive, tangible results. Not surprisingly, however, there was a trade-off between schedule cost and production—improving one objective is, to some extent, detrimental to the other. Fortunately, RAS has automated features for exploring this trade-off.

The metric we use for production is called “production delay”. A low value for production delay is good, because it means less idle time and early drilling of high volume wells (a more detailed definition is provided at the end of this post). In the figure below, we have placed production delay along the horizontal axis, rig cost along the vertical axis, and the client's rig schedule (which we have called the “Manual Schedule”) at the 100% mark on both axes.

When you work on a schedule in RAS, it displays a measure of schedule quality using a variety of metrics. You can experiment with changes to the schedule until the metrics improve. When we make a sequence of changes to the “Manual Schedule” we obtain a series of new schedules with different combinations of cost and production delay relative to the starting point. Clearly, we do not want to end up in the upper right quadrant with higher costs and delay. But, what about the lower right quadrant? It might make sense to increase delay somewhat to achieve lower costs.

Manually evaluating different scenarios and trade-offs can be quite time consuming. But Actenum RAS improves productivity by automating schedule comparison and trade-off analysis.

Which schedules are achievable and how are the trade-offs made? To answer this question, RAS generates trade-off plots. See below for an example of multiple scenarios compared to the original “Manual Schedule”.

For simplicity, we have only shown the manual schedule point and a few others (for how these points were selected, see the notes below). The plot produces a “frontier” of good, feasible schedules (representing different trade-offs).

The plot shows that you can improve production delay and rig cost simultaneously. The points PC, CP, and 5PC are better than the manual schedule for both objectives. Costs have been reduced up to 5% and production delay has been improved up to 40%. In this example, the reduction in rig cost when using 5PC represents an $11.4 million saving.

The plot shows that you can reduce the rig cost further, with 5%, at the cost of almost a 50% increase in production delay. And vice versa, it shows you that you can reduce production delay further by increasing your rig cost. 40PC provides the lowest production delay. The production delay at this point is 38% less than the manual schedule and the rig cost has increased by only 11%.

The figure below shows the production rates for the manual schedule and 40PC over a 36 month period. The integral of these curves represents the cumulative production. As expected, the optimized schedule provides not only a higher cumulative production, but also higher flow rates earlier in time.

Moving production forward in time has a positive effect on Net Present Value. The NPVs relative to the manual schedule for all the scenarios in the trade-off curve are shown in the figure below

In the real world, trade offs are common. The challenge is to evaluate the effects of trade off scenarios on key performance metrics in complex scheduling environments. Actenum RAS supports decision making by simplifying analysis of trade-offs. This benefits users through increased production and lower costs, resulting in higher project NPV.

Notes

• For more detail, see the paper that I presented at the SPE's 2008 Annual Technical Conference and Exhibition in Denver, Colorado. Since then, we have released Version 3.0 of Actenum RAS.
• The “production delay” metric measures the difference between the accumulated oil production of (a) our schedule and (b) a hypothetical schedule using enough rigs to drill all wells on the first day.
• In the trade-off plots above, we only included a few scenarios. These points are said to be Pareto-efficient. A point along the frontier is considered to be Pareto-efficient if there is no other point that is better in both objectives.

I am writing this from Denver, Colorado, where I am presenting a paper at the SPE's Annual Technical Conference and Exhibition, on the worrying decrease in experienced engineers and other knowledge workers in the Oil & Gas industry.

A recent report published by Ernst & Young, in conjunction with Oxford Analytica, gives a snapshot of the ten highest strategic business risks for the oil and gas industry, as identified by the authors after structured consultations with industry leaders and subject matter professionals from around the world. I have constructed the diagram below from a more complex diagram in their report; it illustrates the relative risks of ten major threats to the oil and gas industry:

An increasing human capital deficit is considered the highest strategic risk to the sector. I would add that the human capital deficit we are talking about here is a capability deficit. It is the industry's longest serving and most experienced experts—the knowledge workers—that are retiring in large numbers, causing the so-called “big crew change”.

The result is insufficient personnel with the experience to make autonomous decisions on critical projects and operations in key areas like exploration, development, and production. Exacerbating this is that it is happening at the same time as E&P operations are becoming increasingly challenging. New graduates must solve far more complex recovery problems than the industry has faced previously, and with less experience.

These problems cannot be solved on short notice and may have an effect on oil production. The imbalance in supply and demand of knowledge workers is of such proportions that it is considered a threat to the industry's ability to execute on its planned E&P projects for the next few years, and thus on its ability to meet the demand in the market. In essence, oil field developments may be delayed because of this situation. According to CERA, the shortfall in the industry is already taking hold, with insufficient expertise to meet 2008 exploration and production project demand.

This poses important questions on how to preserve institutional knowledge and increase the productivity of the remaining human capital in order to maintain the industry's capability and productivity levels. The strategies followed by the various industry players involve increased investment in Information and Communication Technology (ICT). Indeed, ICT is recognized as a significant contributor to productivity.

However, while ICT is considered to be a key enabler of performance, it is applied quite unevenly. In particular, automation drops as we move up the information value chain. At the decision level, knowledge workers are still often left to themselves in the last critical piece, the decision making itself. Most decision support tools do not assist in sorting through the many options, weighing the options against each other, and proposing a course of action: they do not have problem solving capabilities. This is unfortunate since it is at the analysis and decision levels that reasoning and industry expertise is in short supply.

The argument I propose in my SPE paper is that the Oil & Gas industry may be able to dampen the effect of the expertise crunch by providing its knowledge workers with software that actively enhances human decision making under complex and changing conditions.

More reading

My colleague Owen Plowman wrote about this topic from an Actenum perspective a month ago. Two articles that are also worth reading on this topic are David Buczek's “Brain Drain: Retaining Intellectual Capital in the Energy Industry” (Journal of Petroleum Technology ISSN 0149-2136, 002, vol. 54, no 1, pp. 26-30) and Christine A. Resler's “Quantifying the Workforce Crisis in Upstream Oil and Gas“ (Talent & Technology, Vol. 1, No 3. 2007).

We're going to hear a lot more about using software that takes an active role in our decision making process.

While software is often used as a tool to help us analyze, it's not frequently trusted to advise us as we take the step from thought to action. But for asset-intensive organizations, exploiting the decision making power of new software applications will separate the winners from the also-rans.

Note that I use the word “trusted” here. Whether we use computers or rely solely on human intuition, we have to “trust” the decision maker. No matter how sophisticated your information collection and analysis capabilities are, you inevitably have to take a leap of faith and actually make a decision based on the conditions around you (using data from the past, present and, to the extent possible, by including predictive information about future scenarios). This leap from analysis to action is where the most value can be gained or lost. This is the point where you put your “intelligence” to the test.

Using technology to go beyond analysis and actively assist decision making is not really that foreign to us. Let's look at a very common example. Driving is something many of us do every day. It involves a lot of complexity, and there are certain skills that we acquire, certain rules that we follow, and different metrics that we use to balance our performance (e.g., journey time, fuel performance) and mitigate risk (e.g., safety, maintenance/repair). We realize that the current systems have limitations. We know that we cannot drive our car looking only at the dashboard. Likewise, we know that we cannot predict an accident by looking in the rear-view mirror. Increasingly, to achieve better performance and mitigate risks more effectively, vehicle manufacturers are using technology to move beyond analysis to provide systems that assist us with complex decision making. ABS, traction control, night vision enhancement, and collision detection are all common examples. We have learned to trust these systems as they help us make complex decisions in changing conditions.

How can we apply the same thinking to management of highly complex production operations?

A January 2007 article in Industry Week focuses on part of this challenge by noting the following limitations:

1. “Most business intelligence looks in the rearview mirror.” Using business intelligence software to analyze past transactions from Enterprise Resource Planning (ERP) software is analogous to driving in reverse along the highway at high speeds.
2. “While real-time dashboards for plant operators [...] have been around for a long time, the data presented to operators have little meaning to supply chain managers, customer service representatives, or the CFO. Real-time manufacturing data need to be put into many different contexts for other roles in the organization that are contributing to the optimal performance of the real-time enterprise.”

The Information Week article addresses the need to use business intelligence discipline on operations data and the requirement to support cross-functional decision making requirements, but more is required. The next step is to go from analysis to active assistance with decision making. Just like the intelligent systems used in our cars, a similar evolution will occur with operations support software. We will learn to trust software that quickly identifies good decisions from literally millions of possible scenarios. Using software to determine “good” decisions will help us achieve business goals and can be measured against key performance indicators. These solutions will assist us to determine what actions to take given rapidly changing conditions and constraints. They will eliminate guesswork, improve performance and mitigate risk. Given the high stakes involved, asset-intensive industries will put more effort into using software that transcends analysis and actually enhances our ability to take effective action.

Now that the Big Crew Change is upon us, and the easy oil is gone, oil and gas producers not only have to focus on rebuilding their base of knowledgeable employees. They also have to maintain existing projects—and develop new ones—while losing the senior people who possess the skills and knowledge to make effective operational decisions in challenging situations.

There’s no quick fix that addresses this situation. But on the positive side, while they are finding and hiring the right mix of people, producers also have the opportunity to rethink and redesign the way that they operate.

More technology isn't enough

An obvious way to mitigate the effects of the Big Crew Change is to use more technology, and to use it in new ways. The various Digital Oilfield initiatives that are in process across the industry are prime examples of this approach. But implementing technology alone isn’t enough: important operational decisions still have to be made on the basis of the flood of information that's generated by much of the technology currently in use.

Capturing knowledge is critical to success

Finding a way to capture the knowledge of experienced employees is critical, to prevent “terminal leakage” of insight into decisions about operational processes, best practices, and so on. Using of optimization technology is one way that producers can capture and apply this knowledge. Actenum's technology, for example, embeds years of cumulative knowledge and experience from multiple individuals into operational models that underlie our applications. So, if you look closely at Actenum MPS, you'll find that it employs such a model to bring together the two worlds of maintenance and production in a way that enables a scheduler with a cursory knowledge of Reliability Centered Maintenance to schedule operations for maximum uptime and throughput, while taking into account the need for preventive maintenance shutdowns.

In an earlier post I mentioned the trade-off between relevance and fidelity of the operational model underlying the optimization application. If the balance is correct, the model will provide powerful insight into decision making for individuals with adequate, but not necessarily expert, levels of knowledge. What's more, this insight can be shared, so that collaborative approaches to decision-making and problem-solving become possible. This has broader implications, since collaboration is often extremely difficult to do well when the experts have informal, unshared knowledge in their heads.

The ball is rolling …

We're starting to see more and more conferences, articles, and papers published on this topic. There's no doubt that producers need to develop strategies to use optimization technologies, and thereby harness the captured knowledge embedded in them, most effectively. This means reviewing operations, looking at where decisions are made, and where gaps in expertise are already appearing or are predicted. It's not hard to find opportunities with immediate (and substantial) payoffs.

Tom Carchrae is the senior engineer working on Actenum's core platform work. He has been leading the development of a new way of applying Actenum's decision and optimization technologies that I find fascinating. It was inspired by the many newspaper stories about ambulances that have taken too long to arrive at the scene of emergency incidents, sometimes with tragic results. The problem is illustrated in a 2005 article in the Philadelphia City Paper: ‘Do you know how many times a week people say to me, “Hey, what the hell, I called you 20 minutes ago?” asked paramedic Lou Rosmini. “But they don't know I just had to drive 50 blocks to get to them.”’

Part of the challenge sits at the emergency control center, where dispatchers are distribute Emergency Medical Services (EMS) ambulances according to unit capacities, meal periods, anticipation of future incidents, and ambulances coming off ER deliveries. Many control centers attempt to deal with the challenge by using predefined deployment plans that will tell the dispatchers in which sequence they should deploy the ambulances to various staging points within a geographic area. These plans include how many ambulances are required for coverage of the area, and where they should be deployed at various times of the day and week.

Building better operational strategies

Actenum has developed a set of technologies and tools that take the guesswork out of developing operational strategies. These tools make it easy to perform predictive analysis (better known as “what-if” analysis) to evaluate different strategies through a detailed simulation that gives you a quantifiable measure of impact. For example, the question “what is the impact on response time of adding an extra unit?” is no longer answered with a vague “better” but rather by a more tangible “improved by 30 seconds”. A more difficult question is “what do I need to do to reach an 8:59 response time?” There are many options to improve response time, such as adding more units, staging locations, altering staging location priorities, staffing schedules, meal policies, types of units, and so on. Actenum's solution searches through the many possible choices to find the best strategy to meet the required targets.

The tool combines call data, road networks, automatic vehicle locator (AVL) data, and deployment policy information, and can operate in two different modes. The first mode runs scenarios over and over while exploring different ways of running your service, while the second determines the best configuration for the ambulance service and proposes recommendations for service improvement.

Dynamic deployment decisions

Setting up the operational strategies and deployment plans is important. However, these plans are often static, and this has obvious drawbacks. For example, they do not take the real traffic or demand load at any given point into account. Experienced dispatchers learn how to make better decisions than the deployment plans would suggest. They try to anticipate incidents and make their decisions based on an understanding of the current situation, which includes consideration of the time of day, weather and traffic conditions, past call history, and population density.

This is, however, not easy, and complexity puts a heavy burden on the dispatchers. Their decisions are neither consistent, nor always very good. They can, therefore, greatly benefit from technologies and tools that support gathering and extracting information, enhance situational awareness, reason with that information, and propose decisions—especially concerning how ambulances should be dynamically deployed to minimize response time.

Actenum therefore applied its dynamic and reactive optimization technology to these challenges. The technology has, in collaboration with CAE Inc., been incorporated into a software application that will be taken to market under the name CAE Deploy. It allows an operator to quickly make plans, easily visualize the planning results, and continue making changes up to the last minute. It is expected to significantly reduce the cognitive load on dispatchers, thereby allowing them to do a better job. Better response times should result, since units are redeployed sooner and more efficiently, and also the paramedic work environment should be improved, because the system considers factors such as meal periods and workload.

Note: CAE Deploy is a result of the work of a consortium consisting of Actenum Corporation, CAE Inc., Precarn Incorporated, Simon Fraser University, McGill University, and Ottawa Paramedic Services.

For most realistic real world situations, optimality is neither achievable nor desirable [1]. Optimality is really a “red herring” in the sense that it has little relevance. Some reasons are:

• Optimal solutions are usually out of reach: A strongly believed conjecture from the area of computational complexity implies that for certain problems optimality cannot be computed easily, even by smart algorithms [2]. Even for modestly sized problems, solutions can't be found in human time. Unfortunately, many realistic industrial optimization challenges belong to this class of problem.
  
  
• Even feasible solutions may be out of reach: Forget about optimality.
  
  
• Optimal solutions are usually brittle: Take, for instance, the problem of minimizing the makespan of a schedule (which is the duration from when the very first task in the schedule starts until the very last ends). If we are allocating our resources smartly we can pack the tasks tightly together (and thereby substantially reduce the makespan). However, such schedules are not robust to changes since there is little or no buffer between the tasks to absorb changes. The moment a task takes longer to finish than planned or a resource shows up too late, then the whole schedule falls apart.
  
  
• Real situations usually have more than one objective: Optimization is conceptually straightforward when you, for instance, want to create an operational schedule that maximizes production. But assume that you want to maximize production and minimize cost. What are you optimizing on? Clearly, these two objectives are pulling in different directions. The more you increase production, the more cost you are likely to incur. Shut down production, and you have minimized cost. The solution is a balance between the two objectives that could be context dependent and can't always be formalized.
  
  
• Optimality is usually not well defined: Usually, companies don't fully know what constitutes an optimal solution. They may recognize good solutions, but it is often hard for them to explain fully why they are good or to give them an absolute ranking. It is a waste of time to chase optimality if optimality is not well defined.
  
  
• Optimality is usually a moving target: In operational environments, it is meaningless to spend an hour or the whole night chasing optimality if the situation changes faster than it takes to propose solutions. What you rather need are good solutions fast.
  
  

Chasing rainbows is good

Optimality is a red herring, but that does not mean that optimization is. Quite the contrary. Chasing rainbows can have great value. Optimality may be out of reach, but we should still strive for finding as good solutions as possible in the time we have available.

For the industrial challenges that we are addressing here at Actenum, there is a diminishing return on chasing optimality. Our philosophy is that it is more important (for many industrial environments) to find good solutions right now than highly optimized solutions tomorrow. “Providing good solutions fast” sounds so simple while being so challenging. Most operational replanning and rescheduling is challenged with cumulative complexities and a huge number of possible combinations. However, if you intend your decision support tool to optimize on key performance indicators in real time or useful time in an operational environment, you have to be clever about which technologies you can apply.

This philosophy brings strong directions to the development of our core technologies. Our optimization algorithms are looking at how we can bring the most “bang for the buck” or value improvement, in the shortest possible time. This is illustrated in the following graph:

The graph shows what happens when we run one of our optimization algorithms for about four minutes (this particular graph is a typical example of a single run of one of the many objective functions in Actenum's rig fleet management system, more concretely the transportation cost minimization function in Actenum RAS 2.0). The graph shows the improvement in the objective function over time when running an optimization algorithm. The x-axis gives time in seconds. The y-axis is the improvement in the objective function. There are several points to notice:

1. It is an anytime algorithm. You can stop the optimization whenever you will, and you will have a solution.
   
   
2. The longer you run it the better solutions you are likely to get.
   
   
3. We get quite strong improvements—25-30% over standard dispatch methods.
   
   
4. The algorithm provides very strong improvements very fast. On these data, we get 5-10% improvement in transportation cost within ten seconds.

Below I have annotated the graph by adding a horizontal line that indicates a hypothetical maximally possible improvement. I have also indicated a point in time where the user has stopped the algorithm (the vertical dashed line).

The question here is really “Does reducing Delta1 have a higher business value than reducing Delta2?”, or in other words, “Where is the best balance between time to find a solution and the value of that solution?”. The answer to this is context dependent, we therefore leave it to the user to decide how long they will run the optimization. Actenum's optimization technologies makes this possible, allowing good solutions fast when that is required.

These methods are not guaranteed to find optimal solutions. However, the methods give you solutions of very high value very fast, and far better than what humans or non-optimization techniques can achieve.

You get something good when you're chasing rainbows.

Notes

[1] In mathematics and computer science, an optimization problem relates to finding a “best solution” from all feasible solutions:

• A feasible solution does not violate any of the declared constraints.
• A best solution is found by ranking the possible solutions, where an objective function determines a solution's rank.
• An objective is what you want to optimize, for instance “minimize transportation cost“. So, a typical objective function might rank solutions according to transportation cost.

[2] Most industrial problems belong to classes of problems that are known as NP-hard. Mathematically, such problems can not be computed in time that is polynomial or better in the size of its input. For many problems, finding optimality may take more time than the estimated remaining time of the universe.

We're working on a proper architectural framework for planning and scheduling in production organizations. It provides a coherent process approach for the various activities and time horizons related to planning and scheduling, and also serves as a basis for the use of our technology solutions.

Why bother, you may ask? After all, everybody knows (or can learn) how to do planning and scheduling, right? Well, what I often find when I am talking to organizations large and small is that there is no architecture or defined framework: the processes are entrenched, and have just evolved over time, to a point where they are difficult to manage, use a mixture of uncoordinated manual methods and software tools, and don't support future growth plans very well.

Below you can see an example (drawn from my discussions with many different organizations) of the kind of situation that I encounter.

The overall production cycle is fed on a yearly basis by production targets, budgets, resource constraints, and so on. Quite often, from that point things become messy. There might be multiple, different, planning processes in place (more than are on this diagram—I haven't shown turnaround planning, for example) which should operate in close collaboration, but which frequently don't (production and maintenance, anyone?). “Planning” may cover a yearly time horizon, or it may look ahead only three to six months. Sometimes there are intermediate scheduling steps which focus on activities between one and three months out, but sometimes there are no such steps. On a daily or weekly basis, there may be all sorts of groups with different processes and different tools, trying to collaborate, with a lot of communication going on (which sometimes gets very fuzzy). I've been in a room at a major production organization where people scheduled by yelling at each other and writing resource/task assignments on pieces of paper. While that might be appropriate in some places, it wasn't working very well for the organization in question.

The architecture that we've devised is shown below.

The overall process is based on the same corporate feeds as in the prior situation, but now we have three major process cycles that are linked and that provide for multiple time horizons: long term planning, mid-range scheduling, and short term operational scheduling and dispatch. Our technology handles all three tiers in this diagram, and we create a centralized database where all information is stored and available to whoever needs it, in whatever form they need it (for example, published on the web in read-only form). Long term planning can be done automatically, and various scenarios for activities and resource allocation can be created and evaluated for suitability. Once a suitable plan is created, it's stored in the database and available to the other processes in a seamless manner. Mid-range scheduling assigns resources to activities in an optimized way, so that organizational targets are met. Again, the results are stored in the database and available to all other parties in the plant. When changes and disruptions occur during the operational day, their impact can be assessed accurately, and various ways of dealing with them can be devised, so that the most appropriate course of action can be put into play.

We think that this architecture makes a great deal of sense. If you have feedback or comments on it, please feel free to send me email.

Technology is a key enabler of performance. Production organizations use a host of software systems to manage their operations: these include various implementations of Manufacturing Execution Systems (MES), Computerized Maintenance Management Systems (CMMS), Enterprise Asset Management (EAM) systems, and Asset Performance Management (APM) systems. At the same time, organizations are seeking ways to take the next step—to extend these technologies to optimize operational performance and automate operational decision-making. Until recently, however, suitable technologies were not available.

That has changed. Developments in decision and optimization technologies now provide us with new advantages. Significantly, they can be used to propose optimized decisions in operational situations, enabling powerful decision support and decision-making. In a presentation that Owen Plowman and I recently gave for the Middle East Chapter of MESA in Al-Khobar, Saudi Arabia, we argued that optimization and decision technologies now have advanced to the point where they provide significant value to MES, CMMS, APM, and EAM systems. Specifically we claimed the following four benefits:

1. You get optimized decisions: Operations management seeks to use MES, CMMS, EAM, and APM applications to provide actionable information—that is abstracted, contextualized, and normalized. For instance, by connecting the control layer with the enterprise layer of information, you obtain new strategic business information and and improved data relevance, and you benefit from advanced dashboards. However, decisions (which actions to take) are still left to the user: these applications don’t propose actions. To go to the next level of sophistication, where proposals for actions are obtained from automated components, we can now use modern decision and optimization technologies.
   
   
2. You turn information into high-quality decisions: In many situations, the applications in use provide you with too much information: a veritable data deluge. Operational signals arrive from sensors, cameras, bar code readers, Global Positioning Systems, backbone transaction systems, RFID systems, vibration sensors, pressure sensors, and so on. But our proficiency at generating information has exceeded our abilities to find, review, understand, and act on it. As a result of the flood of information, our focus has shifted from ending information scarcity to dealing with information overload. Modern decision and optimization technologies turn any amount of information into decisions.
   
   
3. You manage complexity: Operational environments are increasing in complexity, and are subject to increasingly complex operational and environmental constraints. Add to this the volume of information made available from operational systems, and complexity is being driven even higher. As a result, finding the best decisions is often mathematically infeasible. As I said in an earlier post, knowing, at any time, the location of every truck, the pressure at every valve, or the status of every vibration sensor, does not mean that we know how to act on this information. Having a good situational understanding does not mean that we know which decision would be best. Modern decision and optimization technologies are designed to handle complexity. For instance, constraint management technologies remove unnecessary and inconsistent information, while heuristic search methods guide us to promising decisions.
   
   
4. You achieve operational excellence and agility: Increasing operational excellence must be achieved by making optimized decisions, including optimizing the utilization of resources, minimizing waste, maximizing recovery, maximizing production, minimizing production cost, and so on. Increased agility means providing fast and effective decision-making during daily operations, managing disruptions, and responding to events (people calling in sick, special equipment not showing up, weather-related events, …). Many modern decision and optimization technologies are designed to handle non-linear, constrained resource decisions, and recent developments have tackled the issue of doing so in constantly changing environments.

An important part of Actenum’s technology development is focused on providing the world’s best solutions to these four challenges.

After attending the Intelligent Energy Conference in Amsterdam last February, I was struck by Karl Jeffrey’s Letter from the Editor in the April/May 2008 issue of Digital Energy Journal. Much of what he says in “Please can we call this IT again?” is accurate.

There is a lot of confusion around the terms “Digital Oilfield”, “Smart Fields”, and “Fields of the Future”. Understandably, the vision of improving production and operations efficiency using sophisticated data collection, communication, analysis, and decision support tools is compelling. However, there does not appear to be a consensus or model describing the necessary components to support the digital oilfield.

Take, for example, use of the term “optimization”. To some, this means determining the best way to deplete a reservoir. To others, it is a way of automating the use of pumps, valves, and other control systems to ensure reliability and “optimal” flow of product along a system or pipeline. And, to others, it represents a branch of mathematics that can be used to support decision making for challenges in scheduling, planning, blending, and asset management.

Indeed, if you try to use Google to educate yourself about this phenomenon, you will quickly become frustrated. Choosing an effective search term is very difficult. Do you use “oil field optimization”, “intelligent oil field”, “digital oilfield”? Try it and see what you find.

Mr. Jeffrey makes the point that all of this should fall under the rubric of “information technology”. I agree, but with some qualifications.

Information technology (IT) usually connotes a group within a company and not a discipline per se. Its organizational origin dates from the days of the data processing group. Today, the challenges extend far beyond data collection and data management. In turn, the IT skills required go beyond selection of hardware software, systems analysis, application design and development, and systems management. Often missing from typical IT teams are people with domain-specific knowledge (rig scheduling, reservoir planning, etc.) and people with relevant background in decision-making sciences such as operations research and artificial intelligence. These skills are necessary to develop solutions that provide true decision support as opposed to solutions that simply collect, store, and manage data.

Mr. Jeffrey’s suggestion that we should halt the separation of IT from “digital oilfield” groups makes sense. I would simply add that solving “digital oilfield” challenges requires multi-disciplinary teams comprising members with specialized domain knowledge, people with IT skills, and people with appropriate mathematical backgrounds in decision-making sciences. Ideally, these teams should focus on providing actionable operational intelligence to their organizations. What we call it is less important than how we do it, and doing it well requires more than most IT groups can provide.

On May 6th this year Actenum will present at the annual conference for the Canadian Institute of Mining, Metallurgy and Petroleum (CIM). CIM is the leading technical society of professionals in the Canadian minerals, metals, materials, and energy industries, and has over 12,000 national members. It is not surprising, then, that the CIM Conference is the premier mining conference in Canada. The presentation will be held in the "Innovation for Sustainable Future" session chaired by Professor Mike Lipsett, of the Department of Mechanical Engineering, University of Alberta. The session starts at 10:30 am on Tuesday May 6, in Room 11.

Abstract

When you collect data, you’ve simply added cost. You need to add decisions to add value. However, adding decisions about operational situations is challenging. Knowing, at any time point, the location of every truck, the pressure value at every valve, or the status of every vibration sensor, does not mean that you know how to act. Data-rich systems sometimes result in poorer decisions. Decision support systems are usually focused on finding ways of collecting, transmitting, mining, and visualizing data. The last critical piece, decision-making, is left to the user. Frequently, organizations try to overcome this barrier to effective decision making by using various operations analytics tools, including data mining and statistical pattern recognition. However, these approaches provide only a part of the solution. Making decisions, for humans or computers, is far from straightforward. Usually, we have many constraints, competing objectives, and ad hoc knowledge, and often the problems themselves get exponentially more difficult to solve when the number of components increases. Recent developments in scheduling, planning, and optimization have opened new opportunities for improved operations in the resource industry. This presentation will discuss some of these developments and give examples.

You will also find the abstract in the Conference Technical Program under the AM2 Process Improvement track, Room 11, Tue May 6.