Integrated methods of forecasting time series of dynamics of constructing oil and gas wells using additive structural modeling

The paper presents the solution for a problem of forecasting in production systems, which means determining the future values of time series levels, their deviation from actual values being minimized. The construction process of oil and gas wells includes two groups of deterministic homogeneous components which correspond to the intervals of drilling and casing wells, as well as to a component of irregular random impact on the production system proportionate to unproductive time. The time series for the construction of oil and gas wells can be described by an additive structural model, the elements of which are represented by homogeneous components. To forecast work intervals corresponding to zero penetration rates there has been proposed a combined analog-scenario forecasting method. To forecast the intervals of work corresponding to the intervals of drilling wells, a combined method of ensemble-scenario forecasting has been proposed. Scenario forecasting results corresponding to the first quartile, median, and third quartile of statistical distribution are aggregated into an additive structural model of the production system. Evaluation of forecasting effect was made according to MAE and MAPE criteria. The aggregated baseline scenario of the developed methodology demonstrated highest accuracy, the error of which made 12.9%, according to the MAPE criterion, which is better than the results of the standard forecasting method used in operation of drilling enterprises, for which the calculated MAPE value made 14.9%.