A New Neural Network Model for Rock Porosity Prediction

摘要

Artificial neural network has brought a new way for prediction of geological reservoir physical parameters (e.g. porosity, permeability and saturation). However, it becomes strong pertinence and bad universal in parameters prediction. According to the thought of committee machine, the paper presents a new neural network model, which is based on BP neural network, radial basis function (RBF) neural network and support vector regression (SVR) model. And then, a single layer perceptron (SLP) combines different individual neural network to adjust of network structure and reap beneficial advantages of all model. Eventually, a committee neural network (CNN) was constructed. It eliminated the defects of individual neural network in porosity prediction and improved the accuracy of the prediction. Three well logs are applied for experiment. One was used to establish the CNN model, and the other two were employed to assess the reliability of constructed CNN model. Results show that the CNN model performed better than individual neural network model.