Specialists from the Institute of Petroleum Geology and Geophysics SB RAS have improved methods for interpreting seismic data by applying machine learning algorithms. The new technology increases the accuracy of forecasts in the search and development of hydrocarbon deposits, according to the institute's press service.
Seismofacial analysis allows studying underground structures without drilling, using reflected seismic waves. Previously, the interpretation of such data required a lot of labor and manual evaluation. Scientists have introduced clustering and classification methods, including the Bayesian classifier - an algorithm that calculates the probability of an object belonging to a certain type of rock based on already known data.
In the improved version, the algorithm uses not only seismic and well data, but also geological information about the development of reservoirs in the study area. This made it possible to increase the reliability of forecasts and identify promising zones for drilling.
The method has already been tested on a section of a gas condensate field in the Orenburg region. After the analysis, scientists recommended re-perforating a number of wells, and the results confirmed the correctness of the model.
According to researchers, the new technology will allow for more accurate interpretation of seismic data, optimize the exploration and exploitation of deposits, and reduce the amount of inefficient drilling. Further refinement of the algorithm is planned for an even more accurate prediction of the distribution of oil and gas reservoirs.