- 10:35 AM

Multivariable Model Predictive Control Implementation in an Amine Unit at a Syngas (Hydrogen + Co) Plant

Raja Amirthalingam, Chicago Reserach Center, American Air Liquide, 5235 S East Avenue, Countryside, IL 60525 and Omar Germouni, Chicago Research Center, American Air Liquide, 5230 S East Ave, Countryside, IL 60525.

One of the major tasks involved in implementing a Multivariable Model Predictive Control (MVMPC) to any new process is to justify the choice from both technical and economical perspective. Although many multivariable control problems can be best addressed by applying an MVMPC, lack of clear economical benefits can be a serious hurdle in realizing the application. Even if the cost of implementation is justifiable, if there are any symptoms of high maintenance cost that one can predict from the experience of process variations, the idea of implementing MVMPC may not receive enough merit to justify the project. In HyCO plants, Amine system is particularly a very interesting problem to address using MVMPC and also very motivating from the economical perspective. The key challenge in developing a control strategy is to handle a potentially unstable mode that has been repeatedly noticed in our plant at many occasions during normal operations.

This paper compiles the experiences of Model Predictive Control (MPC) application in an Amine System in HyCO plant at an Air Liquide Plant producing hydrogen and carbon monoxide. In addition, we also compare the MPC solution to another control strategy already in place for the same amine system in a different plant.

An amine system which is used for CO2 removal typically contains an absorber for removing CO2 and a stripper to reclaim the solvent to send it back to the absorber. In an amine system, there can be many variations of its behavior from control point of view due to design variations. Often this kind variations is a hard to avoid reality due to the fact that the HyCO plants owned by different organizations undergo many mergers and acquisitions. In general, amine solution is highly corrosive in nature (Dupart et al, 1993) and the problem location can vary significantly from one plant to another and often this problem builds up slowly and at some point suddenly creeps into a serious problem. Some of the examples include, formation of iron carbonates plugging the heat exchanges, corrosion in the recirculation pump impellers, and collapse of the packing due to its mechanical damage, etc. Often, the consequences of equipment corrosion by the amine solvent at high operating temperatures can lead to equipment damages causing high priced maintenance requirement and unexpected plant shutdowns involving production losses. In some design cases, the corrosion losses may be less, but, energy optimization may be critical.

In this paper, the problem that we address for an amine system involves improving the life time of the column packing, improving the column stability, and reducing the operating cost. Although the specified life time of the column packing is 10 years, due to the corrosion problem, the packings were replaced in less than 5 years. In addition to the damage to the packings, the consequences of high corrosion are the formation and deposition of iron carbonates in the amine solution heat exchanges which has many consequential issues to the operation.

The first phase of the control problem involves deciding the appropriate manipulated variables (MVs), constrained or controlled variables (CVs), and disturbance variables (DVs). In the current case, two of the most obvious manipulated variables are amine recirculation rate and the stripping temperature. However, the stripper temperature is a vague definition from control point of view as there are many temperature sensors in the stripper and selection of the right temperature is not a straightforward task. In this paper, we discuss various choices and the issues involved with it.

As the main objective of the amine system is to separate CO2 from the product gas, its concentration in the product gas at the outlet from the system is the key controlled variable. As differential pressures across the column in various locations are key indications of column performance, many differential pressures are also included as CVs. The key disturbance variables would be the total CO¬2 entering the system and the nature of the amine solution itself.

When compared to the MVMPC solution discussed in this paper, another possible control solution is to conduct multiple experiments on the system to model the mass transfer effects to determine optimal operating parameters such as optimal temperature set points, optimal recirculation flow, etc. These parameters have to be determined for various plant loads and operating conditions and the most important requirement is to operate the plant without much modification to the design conditions. This in fact was the case in one of our plants; however, the issue here is the portability. The implementation can typically extend for more time and require more resources and it may not be suitable to all types of plants. In this paper, this comparison is done in more detail and explained.

As one of our plant has been running with the new MVMP control for a year, the experiences from maintenance point of view is also brought forth. This information particularly is aimed to bring out the need for minor maintenance in MVMPC no matter how good they are at the end of first installation.

In order to quantify the original state of corrosion and to measure the improvements, chromium concentration was used as a indicative measure. After the implementation of the new controller, the concentration of chromium has stabilized dramatically and these results will also be discussed in this paper.


1. Dupart, M.S, T. R. Bacon and D. J. Edwards, “Understanding corrosion in gas treating alkanolamine plants”, Hydrocarbon Processing, April-May 1993. 2. Dupart R. S., and R.G.F. Abry, “Amine Plant Troubleshooting and Optimization”, Hydrocarbon processing, April 1995.