Intelligent modelling and optimisation strategies for the BTO process (Bioethanol-to-Olefins)

  1. Sorrosal Yarritu, Gorka
Dirigida por:
  1. Cristina Martín Andonegui Directora
  2. Ainhoa Alonso Vicario Directora

Universidad de defensa: Universidad de Deusto

Fecha de defensa: 25 de junio de 2018

Tribunal:
  1. Gerasimos Lyberatos Presidente/a
  2. Cruz E. Borges Hernández Secretario
  3. Ana Guadalupe Gayubo Cazorla Vocal

Tipo: Tesis

Resumen

Nowadays it is clear that crude oil is an exhaustive raw material and energy feedstock, whose reserves decreases annually. However, it is still one of the main raw material and energy feedstock in the world. This situation makes necessary to develop and industrialise new transformation processes to replace the crude oil as the main raw material. An important technological alternative solution is based on the use of biomass as the alternative to fossil sources. These technologies are grouped in the concept of Biorefinery. This concept includes all the transformation processes, analogous to the petroleum refinery, that are able to produce final marketable products using biomass as feedstock. An important biorefinery process is the transformation process of Bioethanol To Olefins (BTO process). The BTO process, which is the case of the study of the present research work, is a catalytic transformation carried out over an acid zeolite treated catalyst. The BTO process transforms raw material obtained from alternative sources to petroleum, into key products for the petrochemical synthesis, biofuels and plastics like the olefins C3-C4. An important step towards the industrialisation of this type of processes is the development of advanced and optimised control strategies. The main operating variables of the BTO process are the temperature, the water content in the feed and the space-time. Moreover, the more important the effect of the catalyst deactivation becomes the longer is the life of the catalyst. In order to improve the production of olefins and make the process economically profitable, it is important to develop optimised operation strategies for the process. Both the production rate and the catalyst deactivation rate will fluctuate according to the operational conditions. And the fact that both improvement objectives, the production rate and the catalyst lifespan, fluctuate in opposite way depending on the operational conditions, difficult the optimal control strategy. This work is centred in the modelling and optimisation of the BTO process, with the main goal of maximizing the total production of olefins per space-time while at the same time the catalyst lifespan and therefore each production step is prolonged. Both modelling and optimisation methodologies are based on computational intelligence techniques. And therefore, the proposed strategy could be used for any new and unknown chemical processes, properly characterized with a representative set of experimental data, but in early experimental stages, saving costs and time. Soft-computing modelling techniques such as Artificial Neural Networks and Support Vector Machines have been used to model the BTO process. Two modelling approaches have been used, a global modelling and a hybrid strategy which combines soft-modelling techniques with knowledge-based models. In both cases, experimental data of the BTO process and augmented synthetic data generated with a knowledge model of the process have been used during the modelling procedure. The obtained model has been compared with a well-known mechanistic model of the process, which has been used as a contrast method. In order to optimise the operational conditions of the process, three main dynamic optimisation strategies are proposed. The first one is the optimisation of constant set-points, which is the first proposed operation policy, is one the most common control policy in the chemical industry. The second proposed scenario is the operation using optimised fixed-shape temperature trajectories. And the third proposed strategy is the use of dynamically generated trajectories of all the operating variables generated by the optimised Artificial Neural Networks. These three operation strategies have been optimised using evolutionary algorithms as optimisation technique. During the optimisation procedures, the catalyst deactivation is dynamically considered, being its lifespan extension one of the optimisation objectives. The obtained results are compared with the operation of the BTO process with constant set-points optimised at zero time of stream. Finally, the operational conditions are analysed in the last chapter. Different scenarios depending on the cost of the catalyst regeneration phase are presented. The translation of the most suitable operational condition depending on this cost is also studied and presented.