Meet Jim Gilman, a member of SRT's Client Support Team
My background is not in reservoir simulation, in fact my degree is in theatre arts. If you told me 15 years ago I would be working in the energy industry I would never have believed you; actually no one who knew me then would have believed you either!
Jim Gilman is my very patient colleague at Stone Ridge Technology who has sympathetically answered many of my questions on the subject of reservoir simulation software. Jim's 40 year career has included work in unconventional reservoir studies, high pressure/temperature gas, gas condensate recycling, enhanced oil recovery simulation, ranking of geologic models and simulation of naturally fractured reservoirs; and his expertise includes specialization in the application and development of numerical simulators for fluid flow in petroleum reservoirs. Jim worked at Marathon Oil’s Technology Center for over 22 years where he was a co-developer of a 3D, 3-phase simulator for naturally fractured reservoirs; he has authored and co-authored numerous articles dealing primarily with naturally fractured reservoirs, horizontal wells and reservoir simulation. In 2001, along with some associates, Jim was involved in the start-up of iReservoir.com Inc which is where he was working when he first began using ECHELON software in 2015. I recently interviewed Jim who was working from his home in Montana where he lives with his wife. They enjoy the beautiful outdoors there and spending time with their family.
Emily: Can you explain in the simplest way, why E&P companies need reservoir simulators? How you have seen oil companies implement them successfully; and how do the speed and accuracy of the simulator play a part?
Jim: Petroleum companies use reservoir simulation to forecast future production of fluids from petroleum reservoirs. This is for the purpose of optimizing development e.g. how many wells are required and should fluids such as gas or water be injected or re-injected to improve long-term recovery? Over the years, companies have realized that in order to have better models, they need to have a detailed description of the geologic and structural complexity, sufficient gridding to describe the fluid movement and adequate description of the fluid pressure-volume temperature (PVT) behavior. All of this has led to larger and more complex models. However, we never have a unique single description of the reservoir (due to limited information) and so there is a need to address this uncertainty by working with multiple realizations. Additionally, we must validate our models against previous production (e.g. “history matching”). All of this leads to the demand for very fast simulators that can handle many large and complex models.
Just five years ago, when I first started using ECHELON software I ran it on a work station with two GPUs. Now with faster GPUs and more band-width we are routinely running multi-million cell models to honor both complex geology and, in the case of hydraulic fracturing in long laterals, complex fracture patterns. In my career, models have grown from thousands of cells, to hundreds of thousands and now to millions. Multi-million cells are now common and as an example I recently worked on a 49 million cell compositional model for one of SRT's customers, the model was developed to demonstrate the use of water-alternating-gas (WAG) injection to optimize long term oil recovery.
Several years ago SRT demonstrated the ability to solve a synthetic billion-cell black-oil model on a small hardware footprint using a previous generation of GPUs (30 IBM POWER8 nodes each with 4 NVIDIA TESLA P100 GPUs). Run time for this model was about 90 minutes.
Emily: You have used various reservoir simulators in your career; what else is unique about ECHELON other than the fact it has been fully designed and built from inception for GPUs?
Jim: Certainly, the most exciting thing about ECHELON software is the speed and size of models that are enabled by code developed and optimized from the ground-up for GPUs. GPUs are providing dense computing platforms with extremely high memory bandwidth and very efficient arithmetic throughput. This enables a solution for today’s complex and large models in a practical time frame (including compositional fluids).
One unique feature of ECHELON software is the Algebraic Multigrid (AMG) Solver which allows more rapid convergence versus conventional iterative solvers. ECHELON’s AMG solver implementation on GPUs was not trivial, but is now highly optimized for GPUs including the solution of compositional systems. AMG allows a multilevel hierarchy of matrices that can be pre-conditioned to efficiently solve the flow behavior resulting from the underlying complex nature of the geology. In addition to AMG, other unique features are, optimizing flash of compositional fluids, addressing irregular gridding, multi-field networking, ensembles, and uncertainty modeling.
Emily: How important is it to be compatible with other reservoir simulators in the market, particularly Eclipse?
Jim: In a co-authored reservoir simulation primer (“Reservoir Simulation: History Matching and Forecasting, Society of Petroleum Engineers 2013), we talked about reasons for choosing a specific simulator (page 7). This includes robustness, efficiency, ease of use, accuracy, features, familiarity and cost. The main reason users desire that input data files be compatible with existing legacy simulators is to ease the transition for models that are already developed. Additionally, many are reliant on existing software for pre and post processing of input data and output results. For existing long-term users, legacy support is a requirement for easing the transition to a modern simulator.
Emily: What would your advice be to graduates soon entering the engineering industry in oil and gas?
Jim: There will be a long-term need for petroleum products and thus there will be many opportunities for new graduates in the oil and gas industry. The current low-price environment is discouraging, but growth and maintaining today’s production levels requires a knowledgeable and innovative workforce. One aspect of this is reservoir engineering using simulation to address the complex issues discussed previously. Continued development in reservoir simulation software including auxiliary programs for history matching and uncertainty modeling will allow users and scientific programmers to efficiently address the optimization required to economically develop petroleum reservoirs. The next generation of young graduates will be well versed in digital technologies and will thus be ready to tackle these complex issues.
Emily: You have experienced 40 years of reservoir simulation, what do you predict the next 40 will bring?
Jim: The next 40 years will require more innovation, optimization and efficiency to address the economic stress placed on this industry. Of course, a fundamental understanding of reservoirs and fluid flow is and will continue to be very important. For reservoir engineering and geosciences, improved software (in the hands of knowledgeable “subject matter experts”) will be an important part of the solution. Integration of efficient graphical interfaces, with machine learning to handle and interpret large data sets, along with fast full-physics reservoir simulation will allow more effective decision making. More efficient software like ECHELON will aid effective decision making by allowing engineers and geoscientists to efficiently address a wide range of possible outcomes for different geologic models, different recovery mechanisms, alternate development plans and alternate economic scenarios.
So there you have it folks, Jim Gilman on Reservoir Simulation, maybe the next post will be him interviewing me on theatre arts.
Emily Fox is Stone Ridge Technology's Director of Communication.
What we are doing to help improve the reservoir simulation industry.
ECHELON now supports AMD Instinct acceleratorsRead article →
For our first benchmark study on the NVIDIA Hopper architecture, we present ECHELON performance numbers for 10 asset models executing on the NVIDIA Volta, NVIDIA Ampere, and NVIDIA Hopper architectures. The longstanding trend of performance scaling with memory bandwidth continues.Read article →
Leveraging modern GPUs and neural computing libraries, scientists are using deep learning technology to solve differential equations that emerge from a diverse set of physics-based problems. This blog introduces the topic and explores potential applications and limitations of the technology.Read article →