APPLICATION OF MACHINE LEARNING ALGORITHMS IN ESTIMATING RESIDUAL RESOURCE OF FIELD GATHERING SYSTEMS

A model for predictive monitoring of the residual resource and its algorithm are presented, a mathematical apparatus for machine learning systems based on gradient boosting for reliability assessment is proposed, and recommendations for implementation in production are given. It is shown that the model proposed for implementation captures a large number of factors affecting the pipeline, is adaptive and less susceptible to the influence of erroneous values due to data preprocessing mechanisms. A direct comparison of actual data with existing methods for determining the residual resource, as well as with the developed algorithm, is performed, the results of the comparison indicate a higher convergence of the developed model with the actual data. In addition, further possible options for developed models in order to improve the accuracy of the predicted results are reflected.