Блочный метод и искусственная нейронная сеть для обработки и представления данных мониторинга оседаний при нефтедобыче

Abstract

In order to optimize methods of geodetic supporting the monitoring and interpretation data in oil-producing area of a reservoir field, we proposed a modeling method enabling to optimize the construction of a geodetic network and raise the accuracy of determining the earth’s surface deformation using parameters of the model and apply the “block” method for its assessment. The relevance of the block method choice is obvious, its implementation, is to ensure high accuracy of determination and prediction of subsidence. The method enables specifying the re-observation period and dividing the area into parts, which increases the accuracy of the result. The method is effective when using an artificial neural network (ANN). In this case, the ANN consists of two layers, which can be increased in the form of a three-layer network when arranging the forecasting process. At the activation function choice, three similar expressions were considered; the symmetric Gauss function was adopted as the optimal one. In the process of setting up the network for the “block” method, the setting up parameter and the number of inputs (signals) for each individual block for different types of signals were determined.