Application of hybrid artificial intelligent models to predict deliverability of underground natural gas storage sites

Underground natural gas storage is a promising solution to lowering greenhouse gas emissions and attaining sustainable development goals. However, several issues prevent the application of storage projects on a global scale. An accurate estimation of the delivered amount of natural gas from each storage site might be used for supply and demand. Due to this fact, this study proposed hybrid intelligent models integrating the least square support vector machine (LSSVM), differential evolution (DE), imperialist competitive algorithm (ICA), cultural algorithm (CA), teaching learning-based optimization (TLBO), genetic algorithm (GA), and particle swarm optimization (PSO) for approximating the deliverability of underground natural gas storage in different geological formations. We have employed vast data sets of 782 reservoirs from depleted fields to train and validate the proposed intelligent models to predict underground natural gas storage deliverability in the USA. The visual and analytical assessments were used to investigate the performance of the developed intelligent systems. The predicted results showed that all of the intelligent models agreed with the recorded data. Moreover, the statistical indicators revealed that the LSSVM coupling TLBO model shows the highest accuracy in predicting the deliverability of natural gas storage in the depleted field among three intelligent models. Also, the optimal intelligent model accurately predicts 880 and 600 data measurements of saline aquifers and salt domes, respectively. The optimal intelligent model yields a root mean square error (RMSE) value of less than 0.022. The correlation factor (R2) is over 0.998, 0.999, and 0.906 for the depleted field, saline aquifers, and salt domes, respectively. The results highlight the importance of combining smart approaches with nature-inspired strategies in forecasting storage site deliverability. In light of these findings, researchers are better equipped to reduce petroleum energy usage and increase community acceptability of natural gas as part of the transition to green energy.

Automated summary

Underground natural gas storage is a promising solution for reducing greenhouse gas emissions and achieving sustainable development goals. This study proposes hybrid intelligent models to accurately estimate the deliverability of underground natural gas storage in different geological formations. The models were trained and validated using extensive data sets and showed high accuracy in predicting storage deliverability.