(340j) Dynamic Modeling of the Industrial Methanol-to-Olefins (MTO) Process

The methanol-to-olefins (MTO) reaction provides an opportunity for producing basic petrochemicals such as ethylene and propylene from coal, natural gas, and biomass. It sets a bridge between coal or natural gas chemical industry and modern petrochemical industry. Triggered by Mobil’s pioneer work on MTO reaction in 1970s, extensive research contributions since then have been devoted to the reaction principle, catalyst synthesis, and process research and development¹. Honeywell/UOP MTO and Dalian MTO are two main commercial technologies. Since 2010, more than 20 MTO plants have been put into commercial stream.

Promoted by the mainstream commercial catalyst SAPO-34, highly exothermal MTO reactions happen with the reaction temperature around 770K¹. In order to guaranteeing the reaction temperature within the designed region, external cooling system is implemented. On the other hand, coke generated by the reaction distributes on the catalyst surface and subsequently lowers the catalyst activity. Similar to the Fluid catalytic cracking (FCC) process, MTO adopts the circulating fluidized bed reactor-regenerator configuration. Following such configuration, the spent catalysts are recycled to the regenerator where coke deposited on the catalyst is continuously removed. The regenerated catalysts go back to the reactor again. Compared to the kinetic model development and catalyst investigation, building the first-principle model for MTO with the purpose of dynamic optimization and advanced control has received much less attention. To our best knowledge, it is the first time of this paper to touch upon the comprehensive model for the reactor-regenerator system of MTO.

With the aim of dynamic optimization and model predictive control for MTO, the central-elements of this work are to build the first-principle model for both reactor and regenerator and estimate the relevant parameters. Through the careful investigation on our industrial partner’s commercial units, seven-lumped kinetic model for the MTO reaction is adopted². In our work, the whole reaction part is divided into three zones including the mixing region, the transportation region, and the stripping region. The mixing region and the stripping region as CSTRs, while the transportation region is modeled as plug flow reactor. Therefore, they can be described by ordinary differential equations and partial differential equations, respectively. Similarly, we describe the regeneration section as the cascade structure consisting of dense bed and freeboard region. Note that the parameters associated with the reaction section and the regeneration section are estimated individually. Differential algebraic equations (DAEs) are discretized as nonlinear equations by orthogonal collocation finite element(OCFE) method³. The parameter estimation problems are therefore formulated as nonlinear programming. Based on the industrial data sets, we implement the nonlinear programming in the Pyomo model which is then solved by Package Parmest⁴. Comprehensive dynamic simulations are carried out to investigate the performance of the estimated parameters.

Keywords: methanol-to-olefins(MTO), modeling, parameter estimation

Reference:

1. L Ying, et al. A seven-lumped kinetic model for industrial catalyst in DMTO process. Chemical Engineering Research and Design. 2015; 100:179-191.
2. P Tian, et al. Methanol to olefins: from fundamentals to commercialization. ACS Catalysis. 2015; 5: 1922-1938.
3. J Cuthrell, L Biegler. On the optimization of differential-algebraic process systems. AIChE Journal. 1987; 33: 257.
4. https://pyomo.readthedocs.io/en/stable/contributed_packages/parmest/driv....