Predicting Power Consumption of Cryogenic Compressors using Multiple Linear Regression in Machine Learning

Compressor performance is being evaluated based on its power consumption and other operational parameters to meet load demand efficiently while consuming less power. Without proper correlation with other operational data, it is difficult to predict future power consumption that may lead to a low performance of compressors. The Multiple Linear Regression (MLR) analysis in Altair AI Studio software is being used as a model to predict power consumption for four compressors with two different models by considering mass flow rate, suction and discharge temperature, and pressure as its dependent variables. The set of data has been split into two, which are training and testing, at a ratio of 90:10, respectively. This study resulted in a low percentage difference between the predicted and actual power consumption of those four compressors, which are 1.46 %, 1.40 %, 2.00 %, and 2.25 % for Compressor 1, Compressor 2, Compressor 3, and Compressor 4, respectively. The MLR of the compressor power consumption model can be utilized to predict its future power consumption to move towards more sustainable and low-carbon emissions.