Hybrid Deep Learning Approach To Speed Up Reservoir Performance Forecast
The deep learning method has achieved many phenomenal results in image segmentation, speech recognition and even some relatively creative tasks like writing a WIKI page. Intuitively speaking, deep learning can be interpreted as a complex geometric transformation in a high-dimensional space, and the underlying model usually consists of many layers’ representations. Because of the fundamental principle in the design of statistical learning, it normally requires a large amount of data sample to train the model. In many real-world problems, gaining too many data samples is not al-ways practical, (reservoir simulation in the oil and gas industry, for example). In order to overcome this dilemma, we propose to incorporate some physics equation to reduce the amount of data sample needed for training the neural network. In other words, we suggest using prior knowledge in the neural network to accelerate its learning process.
Cheng Zhan is a Senior Data Scientist at Anadarko Petroleum, where he works on field development optimization and long-term production forecasting. He focuses on building machine learning algorithms to create strategic and financial impact for the company. Prior to his current role, he worked as a Geophysicist at TGS and CGG, utilizing seismic data and inversion methods to help operators make better decisions in exploration. He holds a PhD in mathematics from University of Houston, and a B.S. in Mathematics from Sun Yat-sen University.