(343h) Data-Driven Modelling and Optimization of Compressor Operations

Compressors are crucial pressurization equipment commonly used in applications such as air separation, natural gas (NG) processing and liquefaction, petrochemicals, liquefied natural gas (LNG) regasification, etc. They are energy-intensive and usually consume a significant portion of the overall energy expense of a plant.

Generally, multiple stages of compressors mounted on a single shaft (called trains) operate in different configurations such as series or parallel. These trains may operate at different flow capacities such as 25%, 50%, 75%, and 100%. To handle high loads and promote reliability, several trains of compressors exist in a process plant. Starting with the inlet gas conditions and desired pressure ratio, it is not obvious to assign optimal loads in the compressor network. Without a systematic guiding program, operators usually assign unnecessarily high loads to a compressor train and/or choose an inefficient combination of compressors that results in significantly higher power consumption. Therefore, it is important to optimize the compressor operations for varying loads and gas conditions. However, as a preceding step, we must accurately predict the performance of all individual compressors to account for the effects of their geometry, ageing, etc.

Some literature works optimized the compressor operations for specific applications such as LNG regasification. For instance, Shin et al.¹ and Jang et al.² distributed optimal loads to the boil-off gas (BOG) compressors in an LNG regasification plant to minimize their power consumption using a mixed integer linear problem (MILP) formulation. Most importantly, they did not consider stages in a compressor train and various modes of their operations. Besides, they used simplistic compressor models without accurately characterizing the performance of individual compressors.

In this work, we begin with the performance prediction of the reciprocating compressors using operational data. Noting the variation in compressor performance with the gas conditions, we develop correlations to estimate the adiabatic efficiency and constant volumetric efficiency loss. These developed correlations enable the prediction of accurate power consumption and outlet conditions for varying inlet gas conditions.

Then, given the inlet gas conditions and desired outlet pressure, we develop a nonlinear programming (NLP) model to find the configuration and capacities of the compressor trains that minimizes the overall power consumption. Our model results guide the plant operators to assign optimal loads to compressors in the network for varying gas conditions.

As a case study, we illustrate our procedure for the modeling and optimization of compressor operations on a real LNG regasification terminal. We develop correlations to estimate the parameters for the BOG compressors using vendor and operational data. Subsequently, using an accurate predictive model for the compressors, we optimize their operations and achieve significant savings in power consumption.

References:

1. Shin MW, Shin D, Choi SH, Yoon ES, Han C. Optimization of the Operation of Boil-Off Gas Compressors at a Liquified Natural Gas Gasification Plant. Ind Eng Chem Res. 2007/09/01 2007;46(20):6540-6545.
2. Jang N, Shin MW, Choi SH, Yoon ES. Dynamic simulation and optimization of the operation of boil-off gas compressors in a liquefied natural gas gasification plant. Korean Journal of Chemical Engineering. May 01 2011;28(5):1166-1171.