Refinery operations are responsible for a high fraction of
the energy used in the world and have a significant
environmental impact on account of CO2 emissions. One of the
major causes reducing their energy efficiency is fouling, the
deposition of unwanted material over the surface of heat
transfer units. The effects of fouling are more evident in the
preheat train of the crude distillation unit where the thermohydraulic
efficiency can decrease rapidly over time and many
cleaning actions or other control-based mitigations alternatives
have to be implemented. The optimal cleaning scheduling and
optimal control problems are typically addressed separately.
The former has been usually addressed using simple models
and heuristics or stochastic algorithms, due to the complexity of
MINLP formulations with other than unrealistically simple
models. This paper presents a novel formulation and
mathematical programming approach for fouling mitigation
that treats simultaneously the optimal control problem of the
network and the optimal cleaning scheduling, with realistic
dynamic fouling models. The NLP and MINLP optimization
problems are solved via deterministic optimization algorithms.
Using two small examples it is shown that the simultaneous
strategy has the potential to reduce operation cost by more than
10% over and above the use of individual strategies.
Papers presented at the 13th International Conference on Heat Transfer, Fluid Mechanics and Thermodynamics, Portoroz, Slovenia on 17-19 July 2017 .