Control-oriented modeling and simulation of a three-phase gravity separator and its level loop process dynamics identification

A three-phase gravity separator system is one of the most important facilities in the oil and gas industry. The main function of the separator is to separate the three phase inflow of water, oil, and gas into the outlets of the separator. Three phase separators have rich and complex dynamics. Many research activities have been conducted in order to model the behavior of the separator, mostly involving Computational Fluid Dynamics (CFD). Furthermore, few studies also have been conducted from the point of view of flow dynamics and mass balance differential equations.;Control of the three-phase separator processes includes water level, oil level, and the outlet gas pressure. Control of each process is based on the mass balance equations and may correlate with other processes which may affect the overall dynamics of the separator. This issue makes the modeling of the separator a challenging task. A proper model is required in order to control the processes of the separator. The modeling part is done using the mathematical equation of flow dynamics and mass balance of the phases, where each process loop is realized. The model takes into account the tank dimension, the flow rate, the fluid physical properties, and the droplet size distribution. The initial condition is set based on the available data. A simple PI controller is introduced for each process.;The second part in this thesis work will focus on the identification of the dynamics of the level process in the separator. Process dynamics identification is an important problem in the industrial control area. In oil and gas facilities, processes such as level, pressure, flow, and temperature comprise complex unknown dynamics and very often are not easy to control to provide satisfactory performance. To ensure high performance of the designed control, dynamics of such processes needs to be precisely modeled. Process dynamics identification usually consists of selection of an adequate model structure and finding parameters of the selected model. Model structure is selected on the basis of fundamental laws (first principles) used for process description, while model parameters are computed through matching of the actual process and the model responses to certain test signals. The most common type of the test signal is the step input.;It is shown in the present research that the use of the modified relay feedback test (MRFT) as a test allowing one to evaluate dynamics of the process is beneficial to the accuracy of identification. This benefit comes from the fact that identification is done through excitation of frequencies of test oscillations that are most informative and important to the system stability and performance. Data that are measured from MRFT are the frequency and the amplitude of the test oscillations, which makes realization of the test and automatic data measurement easy by means of DCS or PLC. Measurements are done in a few points of the process frequency response (Nyquist plot), and matching of the model response to the actual process response is done through optimization of a certain cost function that characterizes closeness of the measured response and response of the model. The describing function method is used to obtain the frequency response analysis of the MRF oscillations. Another method using the so-called locus of perturbed relay system (LPRS) is also presented in order to perform an exact frequency-domain analysis of the oscillations.;The proposed approach is demonstrated on a liquid level process laboratory setup. The process dynamics is modeled as first-order-plus-dead-time-plus-integrator transfer function. MRFT tests were conducted on the level process laboratory setup having unknown parameters to obtain experimental data. The unknown parameters of the process are found using conventional optimization technique. Simulation and optimization are performed using Matlab/Simulink. The results showed good accuracy of the process identification. A PI controller was then designed based on the process model to verify efficiency of the approach. It is shown that the modified relay feedback test can be successfully used for process identification, with providing a number of advantages over other identification methods.