NewTechnology. Optimizing OptimizatiOn New software technology is revolutionizing reservoir simulation. Into Hot Water

Transcription

1 NewTechnology November 2011 the first word on oilpatch innovation Optimizing OptimizatiOn New software technology is revolutionizing reservoir simulation PUBLICATIONS MAIL AGREEMENT NO Into Hot Water Though a tough sell, geothermal developers making inroads in oil country Water Over Oil Chemical helps produce cold heavy oil too thick to pump at economic rates

2 A new software technology is revolutionizing reservoir simulation By Gordon Cope optimizing optimization The first solution to a reservoir simulation problem is not necessarily the best there may be multiple, equally valid ways of history matching simulation models. Rather than stop at the first viable solution found, optimization will provide a number of versions of the model like parallel universes and determine the optimal of those potential solutions. Images: SPT Group, Schlumberger and Chevron Corporation New Technology Magazine November

3 Every year, finding and producing oil gets more difficult and more expensive. Petroleum companies spend billions of dollars drilling remote prospects and developing new ways of bringing crude to surface, but the only sure bet is that the financial risks are always increasing. Companies spend a lot of money getting reservoir information, and they spend a lot to develop reservoirs, says Greg Walker, a senior reservoir engineer with Talisman Energy Inc., Southeast Asia operations. One of the big questions is always: Do we have a project we can sanction, given the uncertainties? At the other end is the question: Can we make operations as efficient as possible? Over the last several years, a new software tool has emerged that is not only taking some of the uncertainty out, but also improving the chances of success and reducing the workload for geoscientists and engineers. Optimization software costs money, but it s quick and efficient, says Walker. As an example, we were asked to develop well targets for an offshore field. The first time we did it, it took two months to pick two well targets. Using optimization software, the second time it took us one month to pick five well targets. That s a tenfold increase in efficiency, and [that makes things] a lot more interesting for the engineers as they can spend more time on investigation rather than shepherding data through different, separate pieces. Optimization software works hand-in-hand with the simulation software that allows engineers to model reservoir geology, potential oil and gas output, and the most efficient surface facilities, pipeline networks and other infrastructure needed to produce a field. There is a lot of work going on in reservoir optimization right now, says Tony Settari, a professor in the chemical and petroleum engineering department at the University of Calgary. It is very valuable. It gives you more certainty. Simulation software has been around for several decades. Some of the larger oil companies have developed proprietary in-house systems, but the majority of firms use vendor software ranging from simple stand-alone programs to integrated suites marketed by international service companies like Schlumberger Limited and Halliburton. Regardless of the origin, most reservoir simulation modelling tools work in the same manner. They all rely on geological data (such as well core and logs and seismic) that geologists and geophysicists use to create a static reservoir model of the field. The production engineer then uses the reservoir model to create a production plan that will give a profitable net present value to the operator. Information during the initial exploration stages is often scarce, however, so geoscientists and engineers rely on experience and judgment to make reasonable assumptions regarding variables such as permeability, porosity and faulting that might exist between data points. If preliminary production data is available, several history-matching models can be run to compare actual production to what the model predicts. If the fit is poor, the engineer can go back and manually massage the geological model until he achieves a reasonable match. BesT-guess mess Manual-simulation modelling has several shortcomings, however. First, it is very labour-intensive, often taking several weeks to choose reasonable assumptions for field variables. Secondly, even simple models require complex computations that can take dozens of hours to run. As a result, the engineer must weigh time and resources against likely outcome; in the end, relatively few models are run for any one field. The problem is, there are many non-unique solutions to the model, and the engineer may not have found the best one, says David Millar, senior vice-president, reservoir optimization, for SPT Group. So as soon as he gets a reasonable match, the project is done. Oil companies were aware that simulation models gave a best-guess solution BEST OF THE BEST Black spheres on the plots represent possible (local) solutions, but the larger white spheres represent the best (global) solutions. The engineer might find the local solutions and be satisfied, but optimization shows there exist other, better, solutions. with an unknown margin of error, which tended to raise skepticism regarding their value. They wanted a way of reducing risk and uncertainty when it came to planning wells, facilities and pipelines. Building upon academic research at petroleum engineering and mathematics schools around the world, BP plc, Chevron Corporation and other majors began experimenting with probability optimization theorems to improve the quality of simulation modelling. Prior to joining Talisman, Walker spent several years working with BP to develop and deploy Top Down Reservoir Model (TDRM), their proprietary optimization software. Images: SPT Group 16 newtechmagazine.com

4 Image: SPT Group False optima Stagnation region Starting point Global optimal solution For the companies, the attraction is linked to finding new targets, and as a general rule there are a lot more small targets than there are big ones, he notes. In turn, that forces two changes to the workflows; we need a way of getting very accurate predictions and testing a lot of small targets to confirm they exist. Most [manual] production optimization modelling projects six months into the future, and has a per cent uncertainty factor. To get optimization to work, you need to generate more than 100 reservoir models and reduce uncertainty to one to two per cent. Sub-optimal (local) solutions OpTimal SOluTiOnS Engineers tend to stop once they have found just one solution, while optimization can rapidly identify multiple solutions to a range of engineering problems before arriving at a global optimal solution. Optimization relies on a variety of algorithms, or complex mathematical equations; the choice of algorithm depends on the task in hand. A simple optimizer is the gradient method, mimicking the rolling of a marble on an uneven surface to find the nearest low point. More powerful techniques are the genetic algorithm and evolution strategy methods, which mirror the process of evolution where the best of each generation are used to father subsequent generations, says Millar. The best of these children then, in turn, are used to create the next generation, resulting in a steady improvement in the quality of the solution. SPT Group was established in Norway several decades ago to create modelling software that would help predict multi-phase flow from subsea fields. It markets a suite of simulation modelling software that is widely used in the oil and gas sector, including OLGA (multi-phase pipeline and wellbore transient modelling), FORGAS (reservoirs, wells, surface facilities and pipelines) and PIPEFLO (multi-phase steadystate pipeline network design). The company now has over 1,000 customers worldwide and sales approaching US$100 million. In the early 2000s, SPT Group began developing Multi-purpose Environment for Parallel Optimization (MEPO), a software program originally developed for the nuclear Images: SPT Group industry. The company tweaked the program to automatically evaluate a large number of geological variables in order to create a range of potential outcomes. Last year, they launched a commercial version that worked with FORGAS to allow engineers to evaluate a multitude of variables quickly and efficiently in order to create a range of potential production outcomes. Originally, MEPO was used with reservoir modelling, says Mona Trick, advisor, SPT Group Canada Ltd. Last year, we launched a model that can handle any E&P [exploration and production] function, from O&G [oil and gas] reservoirs, to wells and surface facilities and pipelines. AdvAnTAges Optimization benefits oil companies at several levels. First, it allows companies to explore a wide range of reservoir geological scenarios, from the best to the worst, and pick the most favourable ones. This allows you to bracket your space of uncertainty, says Jeffrey Yarus, manager, Earth Modelling, Halliburton Landmark Software and Services (which markets the DecisionSpace suite of interpretive tools). Secondly, it allows quick and efficient matching of historical field data to the simulation model in order to predict future production. History matching can be considered as a bridge between the reservoir modelling and reservoir simulation, says Marko Maucec, research fellow, Halliburton Landmark. Manual history matching is usually performed on a single reservoir model. This will give an idea on how to locally adjust the flow properties of the model, but it doesn t take into account the model uncertainty. Moreover, such manual historymatched and adjusted models are frequently not suitable for reservoir production forecast. A computer-assisted history matching will generate multiple statistically equally probable models in order to capture uncertainty, says Maucec. Because history matching is a mathematical inversion process, it can never give an exact match, but it can minimize the difference of observed and calculated data. You can intelligently select the representative models that account for a reliable reservoir production forecast and capture, as widely as possible, the model uncertainty space. Even with optimization, however, the old adage, Garbage in, garbage out still applies; if geoscientists and engineers start with a bad set of assumptions, they can waste weeks chasing down the wrong alley. To that end, software companies have added preview functions that allow quick computations to see if users are in the ballpark. You don t have to run full physics reservoir simulations in optimization and uncertainty loops all the time, says Dayal Gunasekera, reservoir engineering marketing manager for multiple ways OF looking at THE data Optimization software can quickly examine hundreds of scenarios using multiple data sources to narrow down the best possible outcomes. Schlumberger Information Systems (which markets the Petrel and ECLIPSE suite of simulation modelling software). You can capture the behaviour of the full model by creating a simplified proxy. You run the proxy a few times to get a surface response, then you can run thousands of evaluations of different geological scenarios and dynamic design configurations. This really speeds up the process. New Technology Magazine November

5 uses Optimization has been used in many different upstreamto-midstream situations. For instance, a small gas field in Alberta had four wells producing gas at low pressure. In order to make the field economical, the operator had to choose a production scenario with the best net present value. The most important variables included the number of infill wells, connector pipe size and compressor power. Ordinarily, the production engineering department might spend two weeks running simulation models to find an adequate NPV [net present value], says Trick. In half a day, MEPO analyzed over 100 scenarios and came up with an NPV range with a worst case of $13 million and a best case of $27 million. In another case, an operator was looking to expand a North Sea field, and had done a traditional trial and error reservoir simulation that indicated the field needed six new wells. SPT Group offered the company MEPO on a one-month trial basis. The software quickly determined that the field had a high probability of producing just as much oil using only five new wells. It saved the company US$25 million, says Millar. BP has published a plethora of case studies on TDRM. They note that the typical BP engineer performed an average 300 simulations per year prior to TDRM; each engineer now performs an average of 100,000. The company estimates that the system has added 20 per cent net present value to its reservoirs. Image: Schlumberger Information Services BEST-CaSE SCEnaRiO Steam chambers in SAGD development, designed and optimized with Schlumberger's Petrel and ECLIPSE. BP s Azeri Field in the Caspian Sea had nine steeply stacked reservoirs, some open, some sealed. There were few wells and poor seismic, and huge uncertainty over what production plan to pursue. Traditional fine modelling took 240 hours per iteration. The company used TDRM, and in two days had thousands of simulation cases that showed the best production prospects were sealed reservoirs with well defined gas/oil contacts; engineers were able to focus their production drilling program on that aspect. The Teak Oilfield in Trinidad had been operating for 30 years, and it was time to drill infill wells. Traditional manual history matching showed three locations could add two million barrels each, but TDRM was able to differentiate uncertainty and show well location #2 had the greatest certainty, i.e. least risk. It also showed where the need for more information was highest. It improved certainty and reduced the work cycle by 90 per cent. BP was exploring a carbonate reservoir in the North Sea. Manual history matching gave poor results, due to over 80 variables. TDRM got a good match in less than one month, and showed 50 million to 150 million more barrels present than the manual match. Optimization can also be used on pipelines. OLGA is used all over the world to develop multi-phase offshore and onshore pipeline systems, says Trick. It s very good Images: Schlumberger Information Services REduCing uncertainty A complex well with multilaterals and inflow control devices, designed and optimized with Petrel and ECLIPSE. at determining if problems might arise. MEPO works with OLGA to do history matching of measured pressures and temperatures and optimization, finding the best combination of pipeline size, flow rates and other variables. The engineer can quickly evaluate a range of pipe roughness values, surroundings temperatures, heat transfer coefficients and oil viscosity values to see the effect on predicted upstream pressure. The Shtokman field is located 560 kilometres offshore Russia in the Barents Sea. It contains an estimated 3.8 trillion cubic metres of gas and 37 million tonnes of gas condensate. Since its discovery in the late 1980s, various companies have devised numerous development plans. The current operations group (a joint venture between Gazprom, Total and Statoil) hopes to have the field under production by The plans include a floating production platform and a 560-kilometre two-phase pipeline system. Severe Arctic weather and steep ocean bottom topography add a great deal of complexity to operating the pipeline, however, including hydrate buildup, maximum allowable flow, minimum allowable flow, temperature, pressure and liquid buildup. Shtokman operations group used OLGA and MEPO to run uncertainty analysis on as many variables as possible. They discovered that the most critical parameters to successful operation were pressure drops and fluid buildup. PRoBlems Optimization is not without its shortcomings. One of the dangers of traditional reservoir modelling is that it can give you a false sense of security, says Millar. When you re extrapolating a few wells over many miles, there may not be enough information to make an informed decision using traditional simulation. We were asked by a company that was looking at buying an offshore lease to help them out. There were a few wells drilled, and the traditional simulation model showed that it had potential, but when we ran MEPO for them, they 18 newtechmagazine.com

6 FlOw-THROugH BEnEFiTS Schlumberger used SPT's MEPO to optimize the pipeline and compression system for a small gas field in Alberta. Image: SPT Group realized that even a small percentage less in their assumptions regarding porosity or permeability made it economically untenable. There was simply too little information to rule out that probability. They ended up shooting some seismic, and the new information confirmed poor reservoir quality in some areas. They ended up passing on the deal, and saved themselves a lot of money, Millar says. Industry has taken a pragmatic approach, in that they see how a few types of reservoirs respond to several types of algorithms, and have used that information to experiment with optimization to develop an empirical solution, says Walker. But there are many different types of reservoirs, and the industry may be blind to how those reservoirs are grouped. Companies that aren t familiar with the techniques can easily view the tools as a black box that isn t understood. With the amount of value hanging on decisions, all it takes is one situation in which the technology is used incorrectly to ruin its reputation. More engineers have to learn about optimization, to get an understanding about what the optimization methods can do, and how to use the technology, agrees Trick. To that end, SPT has an academic licence program that distributes MEPO software free to universities with major petroleum engineering departments in North America and Europe. FuTuRe In the near future, optimization will become more and more common, not only because it saves engineers time and improves results, but for financial reasons. In the past, the financial departments in many oil companies have been satisfied with one reasonable answer, says Millar. Now, they are starting to demand to see a range of models, from worst to best, before approving expenditures. Over the next decade, new wrinkles will be added. Geomechanics are not included in optimization, says the University of Calgary s Settari. When you look at field development in the Gulf of Mexico, for instance, your goal is to get as much out as fast as possible. But as the field is produced, it causes reservoir compaction and seismicity along existing faults. You need to take that into account and mitigate it, but it is not reflected in the reservoir optimization. That will be the next horizon they will look at. SPT notes that around 150 customers currently use MEPO and sales are growing strongly. We are very optimistic about the future, says Millar. Currently, optimization is applied to perhaps 10 per cent of O&G modelling scenarios. I foresee that, within a decade, the majority of modelling cases will use optimization. COnTaCT FOR more information David Millar, SPT Group, Tel: , david.millar@sptgroup.com FU. SION n: a merging of diverse elements into a unified whole Page 19 1/2 page horz Cathedral Services Ltd. MWD FUSION MWD SYSTEMS Cathedral s modular FUSION MWD System has taken the industry to a new level by offering leading edge components and capabilities into a single system providing communication, navigating target trajectory and accuracy. While data is being stored to memory, this system provides reliable performance in today s aggressive drilling environment. Drilling + Completions NOW HIRING CathedralEnergyServices.com/careers New Technology Magazine November