Situation
Today Almetyevsk State Petroleum Institute (ASPI) is one of the most respected higher educational institutions of the country that educates talent for the oil industry. In addition to educational activities, the institute conducts applied scientific research, which helps oil producers significantly improve operational efficiency.
The university received a grant from the Federal Target Program No. 05.604.21.0253 "Research and development in the priority areas for the scientific and industrial complex of Russia in 2014–2020" to implement the scientific project "Long-term investment planning for efficient development of oil fields based on high-performance computing and machine learning."
The project involved researchers from the ASPI, Novosibirsk State Technical University, Kazan National Research Technical University named after A.N.Tupolev (KAI), TatNIPIneft Institute, and Tatneft PJSC Business Service Center.
The large-scale project included the research and analysis of modern scientific, technical, regulatory, and methodological literature on the stated topic, as well as patent search for solutions necessary for long-term investment planning. To find a quality solution for the tasks, the researchers needed significant computing resources, which can efficiently process large amounts of data and obtain results in the shortest possible time.
Solution
The NVIDIA DGX-1 hardware and software system offered by Softline met all the requirements of scientists. This supercomputer significantly reduces the time-to-value for high-performance computing and AI projects. With a ready-to-use NVIDIA software stack, the customer can start working with machine learning algorithms in just one day. Moreover, the client does not need to spend time integrating and configuring the necessary infrastructure.
The executives of the oil institute took anticipated that someday the computing capacities may have to be scaled and selected DGX-1. If necessary, the customer can scale out the system's computing resources by building a cluster of DGX-1 modules coupled with an InfiniBand interconnect. Another reason to choose this solution was that NVIDIA offers the NVIDIA GPU Cloud repository—a large library of frameworks and preset neural network models optimized for the GPUs and delivered as Docker containers.
Results
Institute scientists used NVIDIA DGX-1 to select advanced methods, models, and algorithms for long-term investment planning technologies to increase the Tatneft PJSC oil field development efficiency.
"Long-term investment planning for efficient development of oil fields based on high-performance computing and machine learning is a project that constitutes a part of the implementation of the Federal Target Program No. 05.604.21.0253. To implement this project Almetyevsk State Oil Institute received the DGX-1 supercomputer. The system solves many computational problems: generation of development scenarios for Tatneft oil facilities on proxy models; selection of production enhancement operations based on fuzzy neural modeling; generation of optimal oil field development plans in terms of economic and resource constraints. We are grateful to Softline for their all-round support in the project implementation," said Arslan Nasybullin, Head of the Department of Oil and Gas Fields Development and Operation, ASPI.
"Having refined the needs of the university and assessed the scale of the upcoming project, we offered the institute's executives to use the DGX-1 solution. Softline is an authorized partner of NVIDIA for DGX-1 systems that also has in-depth expertise in AI. To ensure the success of the oil university project, we engaged the vendor's experts. NVIDIA experts interviewed customer representatives. As a result of this meeting, we received a confirmation that the tasks faced by Almetyevsk scientists can be solved by DGX-1,"—said Egor Demin, hardware solutions sales manager, Softline.