Prediction of Multivariable Properties of Reservoir Rocks by Using Fuzzy Clustering

Forecasting of geological parameters is very important for decision making on investment to exploration of new hydrocarbon structures and fields. On the one hand, the complexity of this problem is originated from the nonlinearity and uncertainty of behavior of an ensemble of interrelated parameters changing with respect to the depth. This phenomenon is considered analogously to time series, where the depth plays the role of time. On the other hand, the available data are irregular over the depth, as represent different geological bodies with distinct properties. These features mandate necessity to consider multivariable time series of geological parameters with irregular intervals. In this paper, we consider multilag forecasting of five geological parameters over the depth. As a model of forecasting, fuzzy c-means based fuzzy if-then rules are used and this allows better capture of high complexity of the considered phenomena than the classical precise forecasting model. The experimental data show validity of the suggested approach.