METHOD OF NEURAL NETWORK ANALYSIS OF REMOTELY CONTROLLED OIL-FIELD OBJECTS

摘要

FIELD: oil industry.;SUBSTANCE: invention relates to oil producing industry, namely to methods of monitoring state of remotely controlled production and steam-injection wells, submersible equipment at extraction field of ultra-viscous oil (UVO). Method of neural network analysis of state of remotely controlled oil-field objects, which comprises in, that is preparation of data is carried out from the archive of single base containing telemetry data, in the form of n-dimension vectors conditions of wells, which are transmitted to training self-organizing maps by Kohonen, wherein each new state vector for each well is checked for belonging to a specific node using a neural network analysis, additional "critical" n-dimensional vector conditions are introduced, full sets of m from "archive" and "critical" vectors are transmitted to training self-organizing maps by Kohonen, nodes of built Kohonen map are broken into three expert groups based on the obtained groups a statistics of well states are formed.;EFFECT: technical result is a specific method of controlling operation of oil-field facilities and submersible equipment according to telemetry data on extraction fields of UVO.;1 cl, 5 dwg