Seismic Reservoir Oil-Gas Prediction Study Based on Rough Set and RBF Network

In predicting oil-gas reserves in the course of seismic prospecting, a direct employment of neural network will take up enormous storage, lead to prolonged computing time and complicate the structure thanks to the numerous input information dimensions of seismic attributes. To solve the problem, taking into account the characteristics of oil-gas seismic attributes, the paper proposes a prospecting approach based on rough set attribute reduction and radial basis function neural network (RBFNN) . That is to minimize the seismic attributes adopting a rough set reduction algorithm, which will simplify the input structure of neural network as well as cutting down on the time-needed for study and training. The adoption of RBFNN as a prediction system can overcome weaknesses typical of tradition BP network and enable quicker and more stable study and training of the calculation simulated tests and experiments reveal that the network constructed by means of sample attribute reduction is able to meet the precision requirement of prediction and to save costs and enhance processing speed, which has been proved effective in oil-gas prediction