LNG Research Papers - Academia.edu (original) (raw)

With more than 4000 completed plant projects, the Engineering Division of the Linde AG ranks among the leading international plant contractors, with focus on the key market segments olefin plants, natural gas plants, air separation plants, as well as hydrogen and synthesis gas plants. Increased demand and competition for natural resources force the customers of the Linde Engineering Division to improve their efficiency in energy and material utilization.
Due to the need of efficient production, most state-of-the-art plants are highly integrated compounds, optimized for certain design conditions. As the conditions during operation vary from design conditions, adjustments of independent input variables are necessary in order to keep the plants at optimum. This is referred to as optimal disturbance compensation. In highly automated plants, model based controller techniques such as real-time optimization are used to calculate input adjustments for optimal disturbance compensation. However, these techniques suffer from several disadvantages and are thus no standard feature of several plant types such as liquefied natural gas (LNG) liquefaction plants.
In order to nevertheless provide optimal operation, a regulatory control technique was considered in this work. A key concept thereof is that for a particular liquefaction process, there may exist process variable combinations which are almost invariant with respect to optimal disturbance compensation. Thus, the strategy of selecting these process variables as controlled variables and keeping them at fixed setpoints inherently leads to almost optimal operation. Sets of controlled variables which provide inherent optimal operation are commonly referred to as self-optimizing control structures. They can be determined offline using rigorous process models and can then be installed as simple control loops in the regulatory layer. Self-optimizing control offers several advantages over model based control such as simplicity, high operator acceptance, reusability, among others. In general, model based control may achieve a better optimization accuracy than self-optimizing control which makes a combination of both technologies beneficial.
This work aims to develop self-optimizing control structures for LNG liquefaction processes such as the simple SMR cycle, the Linde proprietary LIMUM® cycle and the Linde/Statoil proprietary MFC® process. For sake of high quality process models, the model of a spiral wound heat exchanger, the major equipment of LNG liquefaction processes, was further developed as part of the Linde in-house simulator OPTISIM®. The improvement to predecessor models is the consideration of mass and energy hold-up of streams and the use of highly sophisticated empirical correlations. The model was satisfactorily tested against historical measurement data.
The publicly available methods for the identification of self-optimizing control structures were considered too restrictive as they are only capable of calculating certain structurally limited solutions. Due to this shortcoming, a new identification method was developed. The method manages to find control structures in which each controlled variable can consist of a linear combination of an individual process variable subset with individual set size. Beside the conceptual superiority of the new method, its advance in optimality was proven by a numerical study with randomly generated plant models.
New control structures for the three LNG liquefaction processes mentioned above were calculated by the use of the new identification method. The results were satisfactorily verified by nonlinear steady-state investigations. For both, the LIMUM® cycle and the MFC® process, the new control structures turned out to be economically beneficial compared to conventional control structures. For the sake of integrity, the technical realizability of the new control structure was investigated by dynamical considerations.