1 Modelling, Operation and Control of an LNG Plant Jens Strandberg & Sigurd Skogestad Department of Chemical Engineering, Norwegian University of Science and Technology Trondheim, Norway
2 Outline Statoil's Snøhvit LNG Project Optimal Operation of LNG Plant Contollability Analysis Identifying the Model Results Conclusions and Further Work
3 Statoil's Snøhvit LNG Project Natural gas liquefaction plant situated at Melkøya island outside Hammerfest, northern Norway. Receiving natural gas from the Snøhvit, Albatross and Askeladd fields in the Barents Sea Liquefied Natural Gas (LNG) to be shipped by carriers to markets in Europe and the USA. Plant to go on-line in 2006
4 The Mixed Fluid Cascade (MFC) Process Developed by Linde-Statoil Technology Alliance. Consists of –Precooling section –Liquefaction section –Subcooling section Heat exchangers are plate-fin types and spiral wound heat exchangers. Refrigerants are mixtures of methane, ethane, propane, nitrogen and others LNG product is cooled to -160 º C
5 Optimal Operation of LNG plant Different from Optimal Design What to control? Optimization criteria (Economics) Degrees of Freedom: –4 compressors –4 expansion valves –NG flow –refrigerant compositions (not considered here) –Total: 9 Controllability
6 Optimal Operation of LNG plant Case 1. Given feed rate. Keep final NG temperature at setpoint. Remaining DOF's = 7. Objective function for optimization: However, optimization studies indicate that keeping all the intermediate NG outlet temperatures constant is the best self-optimizing control structure. In this case, the remaining DOF's are 4.
7 Optimal Operation of LNG plant Case 2. Keep final NG temperature at setpoint & Maximize LNG production. NG feed now “free” Compressors optimally at max, remaining DOF's = 4 Keeping intermediate temperatures constant. DOF = 1 So far: only steady-state considerations...
8 Controllability What is the controllability of the plant? Check: –Speed of response to reject disturbances –Speed of response to track reference changes –Input constraints arising from disturbances –Effective time delay Consider one heat exchanger NG Shell
9 Linear Models Controllability analysis -Need linear model of the plant. Starting point: tried to linearize a dynamic model of a spiral wound heat exchanger. Model developed by Sintef Trondheim for Statoil and is a pure simulation model. To derive a linear model by doing perturbations directly on the model equations proved infeasible. Instead, black-box model identification techniques were applied. Matlab's System Identification Toolbox has been used to create low order SISO models for the heat exchanger.
10 Model identification Method applied to the liquefaction heat exchanger. 4 streams –4 input temperatures –4 input pressures –4 input mass flows Simulations made with PRBS- type perturbations. One input at a time. Matlab routines used: –n4sid, pem, bj, oe. NG Shell yu
11 Comparing model outputs
12 Resulting models and Controllability Controllability –speed of responses OK –input constraints OK
13 Conclusions and Further Work Illustrated systematic approach for control system design: –what to control –economics –controllability Usually large difference between optimal design and optimal operation Illustrated use of model identification techniques to derive linear models Detailed economic optimization Linearization of model equations directly MIMO identification Further contollability studies Controller designs Startup optimizations ConclusionsFurther Work
14 Acknowledgements Thanks to Sintef and Statoil for use of Dcoil simulation model Supporters: –Norwegian Research Fund (NFR) –Natural Gas Research Center, NTNU