Development of Statistical Models for Predicting Losses based on the Characteristics of Discontinuities

Abstract

A method for predicting losses over the area of the deposit to minimize the risks of accidents and gas and oil and water showings for the Permian-Carboniferous reservoir of the Usinskoye field was developed. In addition, the analysis of the influence of faults on the number of losses in wells during drilling was carried out. Based on the more than 250 wells drilling analysis, it was revealed that a significant problem during drilling was the loss of drilling fluid. This complication was found in 46% of drilled wells. The intensity of the studied losses was in a wide range: from insignificant losses to strong ones, with a complete loss of mud circulation. The faults identified both from well drilling data and from seismic data were characterized by a different number of wells with and without losses. Using the combination of various statistical methods, individual and complex models for predicting losses in wells depending on the distance from the fault were obtained. Using multilevel probabilistic-statistical modeling, the study of the influence of faults on losses was carried out: initially, based on the data of all wells, regardless of the methods for identifying faults - the first-level model; by the method of identifying faults (drilling / seismic exploration) - second-level models; according to the data of individual faults - the model of the third level. At the fourth level, a complex model was built, which takes into account the calculation results obtained at the previous levels of statistical modeling. The presence of direct and inverse dependences of the absorption probability from the shortest distance to the fault was established. Using linear discriminant analysis, the results of predicting the probability of absorption were checked.