Reservoir Parameters Prediction Using Seismic Waveforms Based on Neural Network

Summary Quantitative prediction of effective reservoir parameters by a set of seismic attributes is an important step in the interpretation of seismic data, and the construction of maps of effective thicknesses is one of the main results of such work and serves as the basis for the laying of new wells in the operation of oil and gas fields. Typically, these constructions are performed based on linear regressions obtained in the process of attribute analysis based on the data of effective thicknesses of previously drilled wells and the selected set of seismic attribute maps. It is known that such an analysis is a rather complex operation and it is easy to obtain an unreliable forecast in complex geological conditions using the wrong set of attributes. Another very common technique is the use of the seismic waveform for seismic facies analysis. Such analysis is usually performed based on unsupervised classification algorithms. It is proposed to use seismic waveform for quantitative forecasting based on neural networks.