Transportation of natural gas

Dynamic Modeling and Control of the PRICO© LNG process

Arjun Singh and Morten Hovd

Department of Engineering Cybernetics

Norwegian University of Science and Technology

Trondheim, Norway

1. INTRODUCTION

For transportation of natural gas (NG), pipeline transportation is often used. However, when gas volumes are moderate, and/or transportation distances are large, the capital and operating costs for pipeline transport become prohibitive. In such cases, transport of Liquefied Natural Gas (LNG) in tankers is often the preferred choice for bringing the gas to the market. In the liquefaction process the natural gas is cooled to around -160°C, and this requires significant amounts of energy. It is therefore important that the process can be operated safely, reliably and efficiently. To achieve this, good control is required.

To understand LNG plant dynamics and to design a robust control system for its operation requires a dynamic model of the plant under consideration. Often the liquefaction unit of the plant is the critical unit which requires maximum attention. To develop a dynamic model for a liquefaction unit of an LNG plant is a challenging task and requires time and effort. The academic work on dynamic LNG plant simulation is limited (Hammer, 2004; Zaim, 2002; Melaaen, 1994). A significant part of these works focuses on modeling of a specific LNG plant. Development in process modeling tools such as Process Systems Enterprise’s gPROMS has made it easier to develop a dynamic model of typical chemical plants such as an LNG plant. This makes it easy to devote significant time to study control aspects in details of such plants. We aim to do so in our work.

The process considered in this work is a single mixed refrigerant process known as PRICO (poly Refrigerant Integrated Cycle Operations) process. (Stebbing and O’Brien, 1975). The PRICO process has been studied from optimization perspective in several publications. (Zaim, 2004, Lee et al., Del Nogal et al., 2005, Jensen and Skogestad, 2006). These works deals with steady state optimization and there is no literature available on dynamic modeling and control structure design for PRICO process. The focus of current paper is to use the model developed for PRICO process (Singh and Hovd, 2006) for control structure and controller design for the PRICO process. In addition to enabling the use of model based tools for control structure design, this allows testing the effects of common model simplifications, such as assuming constant temperature of the refrigerant at the condenser outlet, or ignoring the flash drum and refrigerant holdup. The effects of these model simplifications for model based

control structure development and controller tuning are described in the present work.

2. PROCESS DESCRIPTON

Fig 1 shows the flow sheet of the liquefaction unit of the PRICO process. Some features of the process are removed to make it simple.

Fig. 1: Flow sheet of liquefaction unit of PRICO process

Natural gas enters the heat exchanger with a pressure of around 60 bars and temperature of about 12 C. Natural gas is composed of methane, ethane, propane, n-butane and nitrogen. A mix refrigerant having the same components cools the natural gas in heat exchanger. When leaving the heat exchanger, the temperature of the natural gas has been reduced to around -155 C. The temperature is further lowered to around -163 C when pressure is lowered to near atmospheric.

After compression, the mixed refrigerant is cooled in sea water cooled condenser before it enters the flash drum. After that it is further cooled in the main heat exchanger. The high pressure (~ 30 Bar) sub-cooled refrigerant is throttled in a valve to produce a low temperature two-phase mixture which is vaporized in the main heat exchanger to cool the natural gas and high pressure hot refrigerant. The refrigerant needs to be superheated (by 5-10 C) before it enters the compressor to avoid damage to the compressor.

3. MODELING

A detailed dynamic model for the plant is developed in gPROMS using Multi-flash for calculation of physical properties for the natural gas and the refrigerant (Singh and Hovd, 2006). The SRK equation of state is used for both refrigerant and natural gas. As evident from Fig. 1, first it is essential to develop model for the main components in the plant flow sheet, namely the heat exchanger, valve, compressor, condenser and flash drum. The model of the heat exchanger and condenser are

based on the same principles, the only difference being that in main heat exchanger there is heat exchange between three streams whereas in condenser only two streams exchange heat. Valves are modeled as isenthalpic processes. A brief description of models is given below:

3.1 Main heat exchanger

A one dimensional distributed dynamic mathematical model for a heat exchanger having heat exchange between the three streams is developed using enthalpy and mass balances. Pressure drop in the heat exchanger is neglected. The composition of each stream is assumed to be constant from inlet to outlet. A constant heat transfer coefficient is assumed for each stream. All streams are assumed to exchange heat through one metal wall. The metal wall separating the streams is assumed to have negligible thermal conduction in the axial direction and infinitely fast thermal conduction in the radial direction. A separate energy balance is used for the internal energy of the metal wall. Wall ends are assumed to be adiabatic. The models for both the streams and the wall are one-dimensional.

3.2 Compressor

This model describes the relation between gas mass flow rate and pressure head across the compressor. In this model, infinitely fast dynamics is assumed within the compressor. Negligible hold up and inertia of refrigerant is considered in the compressor. Fan Laws (affinity laws) are used to model speed dependent variations in performance, so that single characteristic curve (head vs. flow) is enough to describe behavior at any speed. The compression process is modeled as polytropic. Constant efficiency is assumed for compressor i.e. efficiency is not assumed to vary with flow rate.

3.3 Flash Drum

It is assumed that the liquid and vapor are at equilibrium at all times and thus there is perfect contact between the vapor and liquid phases. Also it is assumed that there is negligible entrainment of liquid in the vapor stream. The model accounts for the mass balance of each component. The overall energy balance ensures that internal energy is conserved. No heat addition or subtraction is included in the flash calculations, thus resulting in an adiabatic UV flash calculation. This is a standard model from GPROMS model Library.

3.4 Throttling Valve

This valve is assumed to be isenthalpic. Also this

...