МЕТОДИКА ОПЕРАТИВНОГО ПРОГНОЗИРОВАНИЯ ПОКАЗАТЕЛЕЙ НЕФТЕДОБЫЧИ СРЕДСТВАМИ ПРОГРАММНОГО КОМПЛЕКСА С АНАЛИТИЧЕСКИМ МОДУЛЕМ ИСКУССТВЕННЫХ НЕЙРОННЫХ СЕТЕЙ

Abstract

Forecasting and identification is a key element in the overall problem of management. The possibility of its solutions are considered in the application of new information technologies, which are an integral part of intelligent information processing tools. Such means of modern information technologies in the artificial neural network (ANN), the use of which makes it possible to design new hardware and software to significantly expand the classes of tasks and improve the accuracy of their solutions. In this paper we consider an example of calculating the operational oil production by analysis of the АNN. Before calculating, a necessary step is to prepare the input data, the choice of the network architecture, the learning algorithm and its parameters. In this case, the network architecture was chosen, corresponding to the task at hand - a network of direct distribution on the basis of a Multilayer Perceptron and a basic learning algorithm - a method of back-propagation in the form of a batch (group) of the gradient descent algorithm. As an example of the calculation based on the use of ANN benchmark experimental data set of universal modeling technology of hydraulic reservoir pressure maintenance systems Hydraulic Simulator (Hydra'Sym) laboratory software development SunEXe. On the basis of the calculated output data can adequately assess and compare the final results of program complex SoftStatica Neural Analyst.