(443a) Optimal Design of Dynamic Experiments for Pilot Plants for CO2 Capture

The systematic design of dynamic experiments (DoDE), especially for pilot plants for CO₂ capture, is an area that has received little attention in the literature. Since pilot plant test runs require significant resources, Dynamic experiments can easily contain over 1500 data points with meaningful information gathered in a day, while conventional steady-state design of experiments yield 2 or 3 data points. Thus, an optimal DoDE is developed and implemented in a pilot plant for CO₂ capture.

The objective of the DoDE is to maximize the observability of model parameters within a given time constraint for the test runs. For DoDE, it is required that the inputs signals are persistently exicted so that the observability of the model parameters can be ensured. For real-life implementation of the inputs signals, the designed signal should be “plant-friendly”. Plant-friendliness of the input signals ensures that they do not lead to unacceptable changes in product quality and controller set-points that could cause “wear and tear” on the process equipment. Furthermore, the input signals should not lead to unsafe operation of the plant. For designing plant-friendly input signals, one needs to also consider the crest factor, which is the peak amplitude in a signal waveform, in addition to the persistence of excitation. If these properties are not considered, it can lead to signals that are practically unacceptable due to the signal variability, frequency content (harsh changes), amplitude (designed values cannot be achieved at implementation) and waveform (some signal forms may not be implementable in a given control system).

The dynamic model of the pilot plant used for DoDE was developed as part of the U.S. Department of Energy’s Carbon Capture Simulation Initiative (CCSI)^[1]. The model includes rigorous thermodynamic^[2] and transport models^[3] for the MEA-H₂O-CO₂ system, along with a rigorous mass transfer model that was developed by simultaneously estimating the parameters for the interfacial area, reaction kinetics, mass transfer coefficients and diffusivity models by concurrently using the data from wetted wall column and packed column experiments^[4]. A pseudo-random binary sequence (PRBS) is designed as the input signal for DoDE due to its efficiency in obtaining sufficient spectral content^[5]. However, due to the long sequence size, the PRBS signal can be time-consuming and prohibitive for large-order systems. Therefore, a Schroeder-phased input signal, which is a multisine signal, is also designed. This signal has similar characteristics to the PRBS yet can be implemented within reasonable time for higher-order systems. The design also ensured plant friendliness and could be successfully implemented in the Pilot Solvent Test Unit (PSTU) at the National Carbon Capture Center in Wilsonville, Alabama. The transient data are used to solve a dynamic data reconciliation and parameter estimation problem. It is observed that the parameters estimated from dynamic data collected over only a day are nearly equivalent to those obtained from the steady-state data collected over several weeks.

References

1. Miller, D. C., Syamlal, M., Mebane, D. S., Storlie, C., Bhattacharyya, D., Sahinidis, N. V., Sun, X. Carbon capture simulation initiative: a case study in multiscale modeling and new challenges. Annual review of chemical and biomolecular engineering 2014, 5, 301-323,.
2. Morgan, J. C., Chinen, A. S., Omell, B., Bhattacharyya, D., Tong, C., Miller, D.C. Thermodynamic Modeling and Uncertainty Quantification of CO₂-Loaded Aqueous MEA Solutions. Chemical Engineering Science 2017. 168, 309–324
3. Morgan, J. C., Bhattacharyya, D., Tong, C., Miller, D.C. Uncertainty Quantification of Property Models: Methodology and Its Application to CO₂-Loaded Aqueous MEA Solutions. AIChE Journal 2015. 61, 1822-1839.
4. Chinen, A. S., Morgan, J. C., Omell, B., Bhattacharyya, D., Tong, C., Miller, D.C. Development of a Gold-Standard Model for Solvent-Based CO2 Capture. Part 1: Hydraulic and Mass Transfer Models and Their Uncertainty Quantification. Industrial & Engineering Chemistry Research 2018, (submitted)
5. Hjalmarsson, H. (2014). Experiment Design and Identification for Control. Encyclopedia of Systems and Control.