Model Free Optimal and Predictive Control of the K-Spice Process Simulator
Abstract
Impact of the slugging can significantly decrease oil production and have damage consequences for equipment. Nowadays, Model free optimal and predictive control is actual topic of discussion and could be not-exhaustive field for research. The process that describes the behaviour of slug flow was running in K-Spice simulator, while control algorithm was designed in MATLAB.
For implementing the model predictive control with integral action, the state space model of the research process should be known. This model can be obtained by using system identification algorithms e.g. DSR, PEM and N4SID, which could achieve high-performed results with accurate state space model by using only input and output data from a real process. Firstly, it were detected all possible control signals and output variables and then, after detailed review, were formulated four control strategies. For all control strategies were done open-loop simulation for identification the bifurcation point and PRBS experiment for collecting input and output data from the real process that was running in K-Spice simulator.
The process model has been developed by implementing three different system identification methods DSR, PEM and N4SID, which used data from PRBS experiment. The basic algorithm of closed loop system with linear quadratic regulator was firstly discussed and then implemented for each control strategy. The LQR was design based on the obtained DSR model and then tested on the model and implemented to the real process running by K-Spice simulator. In contrast to the linear quadratic regulator was implemented PI controller that was tuned by MATLAB application.
Described procedure was performed for all control strategies, which were highlighted in this project. Achieved results from all control strategies were discussed and it was formulate a list of suggestions for the further work.