Visitors to the 2015 SPE Annual congerence and exhibition in Houston were greeted with a large trompe l’oeil ad for Rock Flow Dynamics’ T-Navigator reservoir simulator even before they had left baggage control. A prelude to the first international RFDyn/T-Nav user meet, held following the ATCE. T-Navigator is positioned as a head-on competitor to Schlumberger’s Eclipse market leader with the plus point that it is tuned for cluster-based use and is cheaper than Eclipse.
Halliburton’s Gustavo Carvajal reported on the use of T-Nav on fine-scale, high definition/high resolution reservoir models of 100 million cells or more. These are now in regular use on large Middle East fields with hundreds of wells and decades of production history. Halliburton benchmarked T-Nav against Eclipse, Nexus and VIP and finally went for T-Nav running on an IBM cluster.
Occidental is a worldwide RFDyn client. Patricia Carreras described the use of T-Nav on a CO2 injection project in West Texas which included a head-for-head comparison with Eclipse. A 60k cell model with equations of state for five fluid components (including CO2) was developed and tested in various production/injection scenarios. The test concluded that ‘T-Nav provided results similar to Eclipse for all development strategies analyzed in this study.’
Martin Sieberer presented OMV’s use of T-Nav to perform multiple simulation runs on a large, mature field for uncertainty analysis. Here a 250k cell model with 315 wells and 60 years of production history was studied at various scales on RFDyn’s Moscow-based cluster. A 9x speed-up was obtained using 8 cluster nodes (320 cores) over the single node, 16 core reference machine. Sieberer concluded that T-Nav offers good compatibility with Eclipse keywords and is easy to use – especially with the Petrel plug-in. T-Nav is also compatible with Emerson/Roxar’s Enable history match tool and provides great scalability. RFDyn’s policy allows a 10x speed up on a 16 core workstation with a single license.
Nice Software’s Francesco Ruffino reported from trials of T-Nav in the cloud, running real time simulation in interactive mode on remote hardware using Amazon’s 3D graphics virtual workstation for remote interaction. The solution worked well from the remote visualization standpoint but Amazon’s clusters only deploy 10 Gig Ethernet, not the faster InfiniBand interconnect. This effectively limits the use for the cloud history match type runs on one or two nodes. Amazon says that if there is enough interest it will the faster hardware in place.
E.On’s Jonathan Carter used T-Nav to perform a large number of simulations (34,000 over a three week period) on a 31 node x 16 core cluster. The statistically-derived final solution was only slightly better than the engineer-designed case but was achieved with much less effort. It is easier to set up 31 models to cover the uncertainty, rather than hold endless meetings to discuss the reference case. Let the computer take the strain! Carter figured that by using Eclipse, the three week project ‘would have needed almost a whole year.’
Wintershall’s Tibor Toth provided another T-Nav/Eclipse head-to-head, investigating T-Nav’s ‘alleged advantages’ of speed-up, cost reduction, intuitive GUI, conversion of Eclipse and CMG data decks and the provision of results comparable to Eclipse. The test was performed by Wintershall’s outsourcing partner ESS on a dual CPU workstation with 20 cores. The result showed that import and export of Eclipse/CMG decks was seamless – although not all keywords are supported as yet. The GUI is ‘excellent and intuitive.’ Black oil and compositional models gave similar or almost identical results to Eclipse. On the downside, a dual phase/dual porosity model failed to run as did a Stars model – probably due to unsupported keywords. Again, speed-up in the 5-7x range was obtained by going from single to 20 cores and significant acceleration was noted over Eclipse.
Denis Zubarev from consultant Amni International Petroleum Resources provided more evidence of the ease of deployment of T-Nav. A 750k cell model with 11 wells and 10 years of history ran in under an hour with ‘no clusters and no license issues.’ Amni’s ‘amalgamated integrated modeling’ approach is claimed to bring an optimal combination of speed, accuracy and resource utilization but, ‘there is no simple and flexible alternative to faster simulation software.’ Use of T-Nav has shown a 4-14x speedup over Eclipse.
Another happy user is Pioneer Natural Resources’ Anthony Quinn who used T-Nav to investigate imbibition processes in fracture systems in the Permian Basin’s Spraberry formation. Pioneer’s high resolution simulations running on a single workstation with dual Xeon X5 processors are as follows: Eclipse (1 core) 7 days, T-Nav (1 core) 13 hours, T-Nav (16 cores) 2 hours – i.e. 84x faster! Quinn concluded that ‘without T-Nav’s hyper-threading this study would have been impractical to complete.’
Tayeb Ayat from California Resources Corp. has used T-Nav to perform an assisted history match for a mature field redevelopment. The project set out to identify new drilling locations, estimate recovery from workover candidates – all in a short, one month time frame. A combination of Schlumberger’s Petrel and T-Nav was used. Ayat reported that Petrel file import to T-Nav was seamless and the assisted history match feature allowed consideration of wide variety of parameters including Corey coefficients. The software was stable and there were no convergence issues. T-Nav proved significantly faster than Eclipse.
While most of the above presentations compare T-Nav with Eclipse, a representative from Petro China unit Xinjiang Oil Co. offered a comparison of T-Nav with Schlumberger’s high-end Intersect simulator. Xinjiang has been using T-Nav for several years and drilled six successful infill wells in 2013 based on the simulator results. But the most interesting trial involved a ten million active cell reservoir model running on a single workstation in Petro China SouthWest. ‘Before we used T-Navigator, it was impossible to run the simulation for the ten million cell models on a workstation. Upscaling was always applied, or we had to run the simulation on clusters. Now we believe that T-Navigator is a much better solution at much less cost.’ T-Nav run time for one model was 3 hours 37 minutes. The model could not run on Eclipse. Intersect, running on a two node 16 core cluster, took 4 hours. Petro China also cited T-Nav’s ‘powerful functionality’ for hydraulic fracturing simulation and concluded that the software was capable of running models with tens of millions of active cells without upscaling and at very low cost in both hardware and software.
KNOC’s Kelly Edwards was also keen on T-Nav’s frac modeling capability particularly as it is completion-centric and does not require special local grid refinement. Also fracs don’t have to align with the grid. Edwards’ study has ‘proved the effectiveness of T-Nav’s frac model.
Tomas Bym introduced Golder Software’s FracMan’s discreet fracture network (DFN) modeling capability which captures seismic and sub-seismic faults. Golder’s workflow was developed for a fractured basement reservoir that proved recalcitrant to simulation. By using FracMan alongside T-Nav it was possible to reproduce the complex observed well behavior.
On the commercial front, Rock Flow Dynamics’ Houston rep Tom Robinson told Oil IT Journal ‘A lot has happened in the past three years. We now have 70 commercial users in fifteen countries including BG Group, Repsol, PDO, YPF and Oxy. Growth has been spectacular - around 50% last year. In PDO, T-Nav has displaced Eclipse and Shell’s in-house developed Modular Reservoir Simulation package MoReS. Elsewhere our price difference is a winner especially when compared with Intersect. We expect 2016 to be another banner year as you can buy a T-Nav license for the cost of maintenance of the current market leader.’ More from Rock Flow Dynamics.
© Oil IT Journal - all rights reserved.