A Multi-Tier Approach for Steam Cracker Revenue Optimization

Optimization models to maximize profits in steam crackers have been implemented widely in the industry. A complete plant simulation including all furnaces, hydrogenation reactors and recovery section results in a highly non-linear problem. Traditional optimization approaches utilize either simplified linear programming models or a steady state plant model in order to reduce the computational costs.

Typically, the optimal process conditions valid for the start of run do not remain optimal for the end of run, owing to the coke deposition, which impact fuel requirements and ultimate plant yield. In order to achieve a real-optimization, it is hence imperative to not only consider the energy costs incurred in the furnace and the compressors, but also to estimate the impact of the dynamic nature of the coke deposition.

This paper describes a novel multi-tier approach [Figure 1] to decompose the dynamic nature of the non-linear problem of steam crackers optimization. The furnaces are simulated rigorously using SABIC in-house cracking kinetics model . At each time-step, the coke-thickness deposited on the cracking tubes is estimated utilizing the rate of coking reaction computed at the previous time-step.

A pseudo-steady state optimization problem is then formulated which is solved in two-tiers.

• In the inner loop, process conditions are generated by optimizing the yield of a simplified recovery section model integrated with the furnace model (Yield Optimizer).
• In the outer loop, the profit is computed using a rigorous recovery section model integrated with the furnace model. Based on the profit difference between two consequent iterations, constraints on the process conditions are relaxed in order to run the yield optimizer recursively.

In this study, our two-tier and dynamic optimization approach is demonstrated via practical example to improve the computational efficiency and to achieve realistic profit-based optimization of olefins plant.