Operational Pigging Prediction by Machine Learning

Abstract

Abstract The paper demonstrates how Operational Pigging frequency is optimized to minimize production deferment and achieve Operational Excellence by predicting the required frequency via machine learning based on the operating condition of the pipeline. Currently, the Operational Pigging frequency is fixed at 3-monthly which leads to 4 times the production deferment required in a year to allow for the Operational Pigging activity to be executed. To perform this prediction, historical data of the pipeline from 2016, which consists of oil flow rate, gas flow rate, running time, water content, and previous operational pigging record,areobtained, and analyzed. From the data, patterns and trends are developed to understand the influence of each data on the prediction model. The data is split into a 60:40 ratio where 60% of the data is used to train the machine learning model, and the remaining 40% of data is used to test the trained model. Based on the historical data obtained, the pigging debris is selected as the target outcome of the machine learning prediction. The trained model using Poison Regression Model upon evaluation shows good accuracy and is able to predict the debris to be collected from the Operational Pigging activity. The prediction of the pigging debris reveals that the Operational Pigging activity frequency can be prolonged fromthe established frequency of 3-monthly to a bigger frequency of as much as 1-year frequency while the debris collected will still be within the acceptable limit. This ultimately allows the operator of oil field ‘K’ in Middle Eastto reduce production deferment as the number of Operational Pigging required in a year is reduced with this prediction. From an Operational Excellence perspective, this prediction model provides a solution to minimize production deferment while ensuring the integrity of the pipeline is not jeopardized.