<![CDATA[Electrical motor driven compressors are among the most energy-intensive components in LNG regasification plants, making accurate power consumption prediction is essential for cost reduction and emission control. Traditional methods, such as Multiple Linear Regression (MLR) are limited by their assumption of linearity, while Artificial Neural Networks (ANNs) offer greater flexibility in modelling nonlinear and dynamic compressor behavior. This study compares MLR and ANN models using real-time data from two Boil-off Gas (BOG) compressors and two Regasification Terminal Export Compressor (RGTEC) compressors. The results show that ANN consistently performs better than MLR. It achieved R² values of 98.3%, 99.9%, 99.9%, and 91.7% for the four compressors. In comparison, MLR reached R² values of 97.1%, 98.5%, 99.7%, and 64.1%. The ANN models also produced lower error magnitudes, including MAE and RMSE. This was especially true under unstable operating conditions when linear models failed to fit properly. Unlike previous studies that relied on simulations or single-method analysis, this research offers one of the first direct comparisons between linear and nonlinear models applied to real-time LNG compressor data. It highlights the practical benefits of ANN for data-driven energy forecasting and optimizing operations in the gas industry. The findings emphasize the value of data-driven methods, particularly neural networks, for improving energy forecasting and operational optimization in the gas sector.]]>
Copyrights © 2026
