Silvya Dewi Rahmawati
Liquid loading is a phenomenon where fluids originally dispersed in gas accumulate due to gas velocity not reaching its critical value. The accumulation in the tubing can cause erratic and slugging flow, leading to a decreased production rate and eventually killing the well. Prediction of gas critical rate was usually done using empirical equations such as (Turner et al., 1969) and (Nosseir et al., 2000). Still, the use of empirical equations requires manual adjustment for different data sets. Therefore, this study presents a more efficient and accurate method of predicting the critical rate of gas well using a machine learning algorithm.
This research uses 94 data from (Turner et al, 1969) and ( Coleman et al, 1991) to build a machine learning model to predict critical rate. From the dataset, variable selections were done by implementing Spearman Correlation. The model selected was XGBoost due to high R2, and low MSE, RMSE, MAE values with low differences in training and testing R2 that minimize the risks of overfitting.< br>
The critical rate predicted with XGBoost was 1.85 MMSCFD while the actual critical rate measured was 1.84 MMSCFD. XGBoost exhibits 0.54% error compared to 7.01% with the Turner Equation, without the use of manual adjustment with a different dataset. Using decline curve analysis to predict the production rate decline of Well-YY and to predict the time of liquid loading, the predicted time was before September 2017, whereas the actual liquid loading time was January 2018.
Collecting gas well data to carry out training for the development of rigid machine learning. Building machine learning to predict the occurrence of liquid loading events in gas wells which can hamper gas production Added machine learning analysis related to well production profile.
The development of this research will enrich the development of machine learning and artificial intelligence related to predictions of liquid loading in gas wells. By implementing this technology, we can avoid dead wells in the field due to liquid loading events and reduce national gas production.