Application Study Of Deep Learning In Reservoir Prediction

Seismic reservoir identification plays an important role in exploration and development of oil and gas reservoirs.With the development of oil and gas exploration,there are higher requirements for seismic reservoir prediction including elastic parameter prediction,physical parameter prediction and lithofacies identification.There are many elastic parameters,physical parameters and rock information in seismic data and logging data.It is extremely important to accurately establish the correspondence between the original massive data and the parameters related to seismic reservoirs.In recent years,deep learning has developed rapidly in the field of machine learning.It has the advantages of deep extraction of large data features and high accuracy of results.The application of deep learning algorithm in geophysics is of great significance for seismic reservoir identification.Firstly,the non-linear correspondence between conventional logging data and seismic shear wave is established by using multi-layer perceptron prediction method.The prediction of seismic shear wave velocity is realized by using conventional logging data,and the prediction accuracy is evaluated by using the correlation coefficient between prediction value and real value.The evaluation results show that the multi-layer perceptron neural network has a good effect in shear wave prediction.Subsequently,this paper classifies the lithofacies of the well location by using the classification function of convolutional neural network,then extracts the seismic data of the well-side channel,establishes the non-linear correspondence between seismic data and lithofacies by using seismic data of sidetrack and rock information interpreted by seismic experts as training data,and applies this relationship to the same work area without wells to classify lithofacies.The classification results show that the convolution neural network has good classification performance.Finally,the porosity is predicted by P-wave velocity,S-wave velocity and density inverted from seismic data.The prediction results of actual data show that the depth neural network has a high accuracy in porosity prediction.The work and conclusion of this paper prove that deep learning method has a far-reaching prospect in the field of geophysics.