Gas-bearing reservoir characterization using an adjusted Parzen probabilistic network

The Parzen probabilistic network which is based on the Bayes theorem was developed using three wells in gas reservoir F3 to estimate porosity distribution. It used seismic attributes of similarity, instantaneous amplitude and energy rather than the conventional post-stack volume. The conditional probabilities involved in the Parzen probabilistic network should be adjusted based on an appropriate smoothing parameter which was calculated by cross validation approach. When the testing dataset was used, total accuracy was obtained as 0.7578 for the developed algorithm. The obtained results indicated that the adjusted Parzen probabilistic network characterized the discontinuities of F3 reservoir as fault structures and gas chimney and also estimated the porosity distribution.