Modelling, estimation and optimization of the methanol synthesis with catalyst deactivation
MetadataShow full item record
This thesis studies dynamic modelling, estimation and optimization of the methanol synthesis with catalyst deactivation. Conversion of natural gas is of special interest in Norway for both economic and political reasons. In June 1997 Statoil opened a methanol plant at Tjeldbergodden. It is among the largest in the word with a capacity of 830 000 metric tons per year, which equals 1/4 of Europe's capacity. Methanol is today mainly used as a building block to produce chemical intermediates, and it is also a promising fuel for fuel cells. The overall theme for this thesis is to find the optimal operation of the methanol synthesis when undergoing catalyst deactivation by dynamic optimization. There has been an increase in method development, tool development and industrial use of dynamic optimization in the past decade. Few realistic, large-scale applications with nonlinear, first principle models, exist apart from this thesis. A rigorous pseudo-steady state model of the total methanol synthesis loop is developed. The model has been verified against a design flow sheet and to some extent, validated against process data. Overall, good agreement was found. Optimal operation policy of the methanol synthesis is found by dynamic optimization. Both the reactor and the reactor system with recycle are studied, and it is shown that the total reactor system with recycle must be considered to find the optimal operating policy. It is also shown that a heterogeneous reactor model gives different optimal policy and more accurate results than a pseudohomogeneous reactor model. Optimization of the loop leads to USD 3 165 000, or 0.75 per cent, increased profit over four years compared to a selected reference case with a constant operation policy. Optimal operation policy is compared with an operating procedure recommended for the Tjeldbergodden methanol plant. The calculated optimal operation policy gives higher profit. However, there are two important advantages of optimization: The ability to find the optimal operation if some of the variables in the optimization problem change, and the ability to track changes in the process by model updating and repeated optimization. During modelling and optimization, it became evident that a good industrialscale deactivation model is needed for the methanol synthesis. The sensitivity in the dynamic optimization of the methanol synthesis is analysed with regards to the deactivation model by a first order error propagation - approach. It is found that 20 per cent standard deviation in the deactivation parameters is suffcient for optimization purposes. A deactivation model for the methanol synthesis catalyst is estimated from historic process data from a methanol plant. A model on the generalized power law form was successfully fitted to process data from a limited period of time. The estimated model is of second order. No measurable effect of water was found, probably because the variations in the feed compositions were too small. The model parameters found are confidential, and are not valid for the total catalyst lifetime because the deactivation process is fast in the beginning and slower after some time. It is necessary to use data from a longer period of time to obtain a model that is valid over the total catalyst lifetime. The work presented in this thesis serves as a framework for the implementation of dynamic optimization in the control system of a methanol synthesis plant. The dynamic optimization should be implemented in the optimization layer of the control system with feedback to update the activity. A new catalyst deactivation model should be estimated whenever the commercial catalyst type is changed. A shrinking horizon algorithm with repeated optimization is proposed. The main contribution in this thesis is a realistic, large-scale case study on modelling, estimation and dynamic optimization. Several researchers have studied optimal operation of fixed bed reactors experiencing catalyst deactivation. This work adopts an approach which is more realistic. A rigorous model of the total reactor system with recycle, with varying model parameters and thermodynamic properties is used. A thorough analysis of the process is enployed to formulate the optimization problem. The actual time varying control variables in the reactor system, the recycle rate and coolant temperature, are optimized with regards to an economic objective, and path constrains on the reactor temperature are considered. An optimal operation strategy for the methanol synthesis has not been published before. Similar studies have probably been performed in industry without being published.