Online PID Tuning for Unstable Processes: Application to Anti-Slug Control
Master thesis
Permanent lenke
http://hdl.handle.net/11250/248564Utgivelsesdato
2013Metadata
Vis full innførselSamlinger
Sammendrag
Slug flow is an undesired multiphase flow phenomena causing problems in the processing of oil and gas. Anti-slug control is therefore an important research topic. Satisfactory results have been achieved in anti-slug control by use of the simple PID controller. However, as the slug flow system is a highly nonlinear process online readjustment of the controller gain is required to counteract the nonlinearity of the system. The slug flow system is also characterized as an unstable system and most of the online PID tuning rules are designed for stable systems. The first step on the way to find good tuning values is to obtain a model of the process. In this thesis closed-loop identification techniques were applied to different slug flow systems. A set point step change technique, an optimized curve adaptation technique and a relay feedback identification technique were tested. The identification techniques were applied to simulations in Matlab, simulations in the multiphase flow simulator OLGA and on a small scale laboratory set-up. Three sets of analytical PID tuning rules designed for unstable systems were applied to the systems under investigation. All of these tuning rules contained a tuning parameter, $\lambda$, which was optimized for performance and robustness. The three sets of tuning rules were constructed for three different PID controller structures. Optimal PID tuning parameters were then found for each of these structures. The optimal tuning rules were applied to Matlab simulations and a small scale laboratory set up.It was found that all three identification routines performed well. The optimized curve adaptation method gave the best identification of the process in all cases, but the routine did not manage to identify the time delay exact. However, it did identify the time delay well enough for most purposes. Optimal tuning values were found to be too aggressive using the Cho et. al. and Chidambaram et. al. tuning rules. The best tuning results, with optimal performance and robustness and small controller usage was found when using the IMC based tuning rules developed by Jahanshahi and Skogestad. Recommendation to further work is to apply the closed loop identification techniques over a larger operational interval. In addition the identification routines should to be tested on different unstable systems, as well as on systems with a different structure than the slug flow system. It is also suggested to conduct more investigations of the optimizations on the PID tuning rules for unstable models.