Accelerating and Visualizing Machine Learning for Reservoir Simulation

点击翻译按钮获取中文摘要

Artificial intelligence (AI) techniques have been identified as having great potential to provide value to the petroleum industry due to the complexities of managing, processing and interpreting vast amounts of available reservoir geophysical, geological and production data. Machine learning, in particular, is on a rapid ascent enabled by the growth of advanced computing power. In the past several years, volatility in the price of oil has driven petroleum producers and service companies to strive for higher energy efficiency, lower operational costs, and less unscheduled downtime. To achieve these objectives, many technical challenges related to the development of machine learning techniques in reservoir simulations must be addressed with a comprehensive research program. In our industrial research chair program, we are investigating many of these challenges. As artificial intelligence and machine learning techniques are used to enhance Reservoir Engineering workflows, it is important to understand how these techniques may be enhanced through the use of acceleration techniques, both traditional and emerging, to mitigate the computational challenges. This enables faster results and more rigorous simulation studies. It is also important to understand how visualization and interaction with the machine learning algorithms may help produce more accurate results and provide greater insight into complex reservoir engineering scenarios. The objective of this research project is to investigate the development of acceleration and visualization enhancements to improve and accelerate machine learning techniques applied to reservoir simulations. The outcome will advance machine learning algorithms, quantum algorithms, visualization techniques, hardware acceleration techniques, history matching, and optimization techniques used in the simulations. Techniques will be developed that directly support the enhancement of the machine learning components of our novel reservoir simulator.