Major countries are installing Liquefied Natural Gas (LNG) terminals worldwide, as they transition towards carbon-free economies. Compressors are energy-intensive equipment in LNG import/export terminals. While reciprocating compressors are in wide use, models to estimate volumetric and energetic efficiencies do not exist, especially for those with suction valve unloaders. Furthermore, commercial process simulators such as Aspen HYSYS or Unisim are not equipped to simulate them rigorously. This paper presents a procedure to develop empirical models for predicting flow and power based on process insights and real operational data. It also demonstrates how these models can be embedded inside simulators to simulate compressor operations in both steady and dynamic modes. Real data from a BOG compressor train in an LNG regasification terminal are used to illustrate the full range of their applications. Finally, the suitability and efficacy of data-driven machine learning approaches are evaluated to show the superiority of proposed empirical models. (C) 2019 Elsevier Ltd. All rights reserved.