Well monitoring can provide a continuous record of flow rate and pressure, which gives us rich information about the reservoir and makes well data a valuable source for reservoir analysis. Recently, it has been shown that machine learning is a promising tool to interpret well transient data. Such methods can be used to denoise and deconvolve the pressure signal efficiently and recover the full reservoir behavior. The machine learning framework has also been extended to multiwell testing and flow rate reconstruction. Multiwell data can be formulated into machine learning algorithms using a feature-coefficient-target model. The reservoir model can then be revealed by predicting the pressure corresponding to a simple rate history with the trained model. Flow rate reconstruction aims at estimating any missing flow rate history by using available pressure history. This is a very useful capability in practical applications in which individual well rates are not recorded continuously. The success of rate reconstruction modeling also illustrates the adaptability of machine learning to different kinds of reservoir modeling, by adjusting features and targets. Machine learning is also a particularly promising technique for analysis of data from permanent downhole gauges (PDG), given that the massive volumes of data are otherwise hard to interpret using conventional interpretation methodologies.
Roland N. Horne is the Thomas Davies Barrow Professor of Earth Sciences at Stanford University, and Professor of Energy Resources Engineering. He was Chairman of the Department of Petroleum Engineering at Stanford University from 1995 to 2006. He is an Honorary Member of SPE, and a member of the US National Academy of Engineering.