Equipment Sizing and Costing

Based on the simulation results, it is possible to perform a preliminary equip-ment sizing that is sucient to estimate their cost to the required precision.

This achieved by using the Activated Economics in Aspen Plus. This is an auto-mated interface between Aspen Plus and the software Aspen Process Economic Analyzer^® (APEA), which contains sub-routines for sizing and an up-to-date cost data bank of modular industrial process equipment. This modules include the equipment manufacturing itself, so as painting/insulation, installation, and the peripheral piping and instrumentation.

Running activated economics is, unfortunately, time-consuming and the re-sults are not accessible by other blocks, e.g., optimization block, nor by external software. Hence, the approach is unsuitable for optimization. Therefore, equip-ment cost correlations have been regressed based on data sets created manually with APEA. The selected cost drivers for each equipment type are summarized in Table 2.5 and the correlation for packed columns is given by Eq. 2.30 and presented in Figure 2.16. The custom model for the membrane module also includes correlations for its sizing and costing. All correlations are provided in Appendix E.1. They are implemented in Python and used in the formulation of the objective functions for the optimizations in Chapter 4.

Table 2.5: Cost drivers selected for equipment cost correlations Equipment Type Cost Driver 1 Cost Driver 2

Column Diameter Packed Height

Compressor Work / Duty

-Drum Diameter Height

Heat Exchanger Area

-Pump Work / Duty

-Cost_column= 72483 + 123075·d^1.10_column·h^0.179_pack (2.30)

Figure 2.16: Installed equipment cost of packed columns. Cost estimates by APEA (dots) and by correlation (surface)

3 Optimization Methods

This chapter details the methodologies used to carry out the optimal design of the industrial-scale BG-OCM plant. The goal is to maximize product yield and minimize operational and capital expenditures, which is achieved in Chapter 4.

The developed Aspen Plus models run in Sequential-Modular (SM) mode, thus suitable methods using stochastic and Surrogate-Assisted Optimization (SAO) algorithms have been developed and applied.

Optimization is essential for process and product development and should be applied from an early development stage. When comparing dierent processes and products, a sub-optimal design may wrongly lead to a candidate solution being discarded simply because information on its full potential was unavailable (Esche, 2015). During the process development timeline, it is thus essential to apply optimization techniques to the process synthesis and process design.

In the process synthesis stage, several alternatives of reaction and separa-tion systems must be evaluated and compared, which is typically achieved by heuristics, superstructure optimization, or hybrid methodologies (Barnicki and Siirola, 2004). For superstructure optimization, the large number of binary de-cision variables often requires a reduction in the model complexity to keep the solution attainable. A novel approach is the utilization of phenomena-based rather than equipment-based blocks, which enables process intensication op-tions to be considered at this stage (Kuhlmann et al., 2018). The next step often consists in developing more rigorous models to be used for a conceptual design and evaluation of a few of the best performing process structures. The optimization carried out at the conceptual design stage usually contains fewer binary and more continuous decision variables, i.e., the optimization is carried out once most of the integer decisions have been made. Hereinafter this is referred to as optimal process design.

Denition: In the context of this thesis, optimal process design is dened as the owsheet structure/conguration, i.e., set of unit operations and their connectivity and specications, and the operation conditions, i.e., ow rates, pressures, and temperatures, that minimizes a given objective function, e.g., total annualized cost, subject to equality constraints, e.g., mass balances, and inequality constraints, e.g., required product purity.

3.1 Previous Work

In previous work carried out within the research group of Process Dynamics and Operations at the TUB, Salerno-Paredes, 2012 analyzed dierent integration options for the OCM process with co-generation, formaldehyde, or oxygenates production. Godini et al., 2013 evaluated the techno-economic feasibility of integrating OCM with dry reforming of methane and Spallina et al., 2017 com-pared OCM to naphtha steam cracking, considering dierent OCM reactor options of increasing complexity and performance. More recently, Godini et al., 2019 also used mini plant experimental results and simulation models to assess dierent integrated OCM congurations on an industrial scale. All of these studies relied on SM simulation models, e.g., in Aspen Plus or HYSYS, and do not seem to have performed optimizations, likely due to the diculties associated with SM optimization discussed in Section 3.2.

In Esche et al., 2015, a superstructure optimization problem for an OCM mini-plant has been proposed. The main focus and methodological contribu-tion therein lie in the incorporacontribu-tion of model uncertainties into the optimizacontribu-tion problem via chance constraints, but some compromises had to be made to re-duce the size and complexity of the problem. Only the ve main components have been considered, i.e., CH₄, O₂, N₂, CO₂, C₂H₄. The H₂O generated in the OCM reaction can be easily condensed out of the main process stream, but CO and H₂ are also present and aect the downstream separations. Phys-ical properties have also been simplied to some extent by the use of non-phenomenological models to facilitate numerical calculations. The distillation section has not been contemplated, hence a full picture of the process is still missing. Finally, the objective therein is to minimize the separation energy re-quirements for the mini plant operation. To enable the optimal process design, however, the xed capital investment must also be considered.

Hence, this thesis complements previous studies by developing rigorous mod-els and a framework that can be used for optimal process design. The modmod-els described in Chapter 2 are phenomenological and allow for sizing and costing of process equipment. The framework described in Chapter 3.3 enables process design based on them.