Efficiency optimization of a centrifugal compressor driven by a steam turbine through machine learning

The study's primary purpose is to explore the potential of digital initiatives in the oil and gas industry by developing a viable and applicable product with a data-driven perspective. Aligned with this context, this project will expose an optimization of the efficiency of a compressor driven by a turbine through Machine Learning. The algorithm will give solutions that positively impact equipment efficiency. The development covers understanding the rotating set, its respective mechanical and thermodynamic analysis, and the selection of the Machine Learning algorithm. The rotating assembly consists of a four stages propylene centrifugal compressor and a extraction steam turbine. Regarding the thermodynamic analysis, the polytropic efficiency of the compressor will be calculated, and for the turbine, the calculation will be based on this isentropic efficiency. For the selection of the algorithm, the programming logic must be considered. In historical mapping, XGBoost will be used as it is an appropriate algorithm for supervised and categorical Machine Learning. The trend analysis of historical performance is conducted for each piece of equipment to better understand its influences and impacts. A modification in pressure conditions is proposed for efficiency optimization to enhance its operational conditions and the efficiency gradient. For the Steam Turbine, the results from this study showed the Extraction Pressure as the variable that most influences the equipment performance. For the Centrifugal Compressor, the variables from the second stage were the most influential ones. For both rotating equipment variables, respective modifications in 5% have an impact on modification in the efficiency category98 p