Ranking of geostatistical reservoir models and uncertainty assessment using real-time pressure data

This paper presents a novel approach to analyze the quasi-continuous pressure data for ranking high-resolution geostatistical reservoir models and uncertainty assessment. Real time monitoring of pressures through permanent downhole gauges is a recent development. A robust procedure to effectively use the enormous amount of data recorded by theses monitoring systems has been proposed and tested on a synthetic case. Geostatistical simulations involve generating multiple equi-probable fine scale depictions of the reservoir heterogeneity each honoring the data available. The simulated pressure responses from these realizations could be quite different, yet the responses are not completely random. In spite of their differences, there are patterns, which occur in theses simulated responses. Such patterns can be identified by means of a mathematical tool called Principal Component Analysis. The classical face recognition technique is then used to rank the geostatistical reservoir models. The simulated pressure data from the multiple realizations is analogous to "training set of faces" while the recorded or the historical data is the "face", which needs to be recognized form the training set. The method attempts to identify the geostatistical reservoir models, which show reasonable match in the dominating patterns in the simulated pressure data with the recorded pressure data. This approach mimics the "face recognition" or the 'Voice recognition" technique, which are already being successfully applied in their respective domains.