System and method of predicting reservoir properties

Abstract

A system and method for predicting oil reservoir properties throughout the reservoir using well data and seismic data. While the reservoir properties in the vicinity of a well borehole are usually known (i.e. by well logging), the reservoir properties between wells are unknown and difficult to estimate. Because the relationships between reservoir properties and seismic data is seldom obvious, the method uses Artificial Neural Networks (ANN's) to estimate these relationships. First, the method calculates the intersections between the seismic data and wellbore date, i.e. seismic reflectors are correlated to geological markers in the wellbores. Second, the method estimates the significance between seismic attributes and borehole properties. Next, the method models or calibrates the relationship between reservoir properties and seismic attributes by training an ANN using seismic attributes and wellbore data close to the intersections. Finally, the trained ANN uses the seismic attributes to predict the reservoir properties between wells. Advantageously, statistic methods are used to estimate the confidence of the predictions.