ВЫБОР ПЕРЕМЕННЫХ ДЛЯ ОБУЧЕНИЯ НЕЙРОННОЙ СЕТИ ПРИ ПРОГНОЗИРОВАНИИ ЭФФЕКТИВНОСТИ ГИДРАВЛИЧЕСКОГО РАЗРЫВА ПЛАСТА

Abstract

Every year more and more different researches of geological oilfield objects and processes which take place in productive reservoir and ground infrastructure nodes are realized with information and computer technologies at all stages of development. We can hear increasingly the term «Intellectual field development» with respect to which the complex models become particularly significant. These models are applicable to production data management system, system of planning and assessment of efficiency of geological and technical measures (GTM), adaptive control system of all other production processes. Proper planning and assessment of efficiency of GTM allow to design the most appropriate and flexible development system from the beginning of development process and consider the possible risks associated with the uniqueness of each operated object. Later it will help to reduce the rate of production decline and achieve project oil recovery factor (ORF). For effective planning and assessment of different types of GTM the combination of geological and production analysis (GPA), based on the collected experience of developing of specific operated object, as well as on the foundations of the general development, and wide opportunities of computer methods in crating of different types of researched object models is used. This article discusses the problem of choosing the input and output variables for the neural network training to predict the efficiency of formation hydraulic fracturing. Neural networks are a type of imitation modeling (IM) and should be interesting for design engineers from practical side because neural networks can be operatively used to obtain the satisfactory predictions, and their using does not require hydrodynamic modeling skills from specialists.