Application of neural networks of long short-term memory for forecasting paraffin deposition processes in main oil pipelines
Abstract
Application of neural networks of long short-term memory for forecasting paraffin deposition processes in main oil pipelines
Incoming article date: 21.02.2025The article studies the application of neural networks with long short-term memory (LSTM) for forecasting the precipitation of asphalt-resin-paraffin deposits (ARPD) during oil pumping through main oil pipelines. The authors of the article outline the relevance of the problem of ARPD formation in main oil transportation and consider modern approaches to mathematical modeling of forecasting the precipitation of deposits. The aim of the study was to develop a neural network model that allows constructing a graph of the distribution of ARPD along the length of the model pipeline over time. Taking into account the features of various types of neural networks and the available input data, a corresponding neural network model based on LSTM was developed. The key parameters of the "oil - pipeline - soil" system were determined, which should be taken into account as initial data. The developed model demonstrates a sufficient degree of forecasting accuracy and at the same time has prospects for its improvement. The results obtained can be applied by operators of main oil transportation for more accurate forecasting and determining the most cost-effective period for cleaning the pipeline.
Keywords: recurrent neural network, asphalt-resin-paraffin deposits, neural networks, forecasting, short-term long-term memory networks, oil trunk pipeline, oil transportation