CENTRIFUGAL COMPRESSOR DESIGN AND SURGE SIMULATION FOR ACTIVE INFERENCE BASED CONTROL

Among myriad industrial applications, the centrifugal compressor is a critical equipment in both upstream and downstream oil and gas production and processing facilities. In upstream applications, the varying properties (temperature, pressure, mass flow) of crude oil during the life of a reservoir coupled with different modes of operation (crude composition) emphasize compressor surge as a crucial factor in its design and operation. The surge constrains design leading to oversizing and higher capital costs. It also narrows the safe operating range during service, resulting in suboptimal performance and higher operating costs. With its origins in neuroscience, active inference (AIF) is gaining traction in industrial applications involving sequential decision making under uncertain and noisy conditions. This paper proposes a novel AIF based advanced process control methodology for robust surge control under uncertain and noisy conditions. AIF based control offers several advantages over traditional methods like Proportional-Integral-Derivative (PID) and Model Predictive Control (MPC). It inherently integrates perception and action, allowing for more adaptive and context-aware decision-making in uncertain environments. As a prelude to building an AIF surge controller, a natural gas centrifugal compressor is designed for a set of input specifications using a through flow model to generate characteristic or performance maps. The maps are then embedded in an ASPEN HYSYS (R) dynamic model to simulate operational conditions that result in surge. The key contributions of this paper are designing a natural centrifugal compressor prone to surge conditions due to highly uncertain suction conditions, integrating the compressor maps into a dynamic system model and simulating surge scenarios, and developing a methodology for AIF based surge control.