Research On Refinery Planning Optimization Based On Genetic Algorithms

Refinery planning optimization has attracted more and more researchers, with improvement of petrochemical industry management. However, most researchers paid attention in model adjustment and dimensionality reduction while works has not been reported on systematic research of optimization algorithm. Therefore, some related work about the optimization of nonlinear and uncertain problems is carried out after the major issues on refinery planning optimization are summarized. Meanwhile, an integrated optimization platform is developed. The main contents of this dissertation are as follows:In Chapter 1, the overview of petrochemical industry and the basic concepts of refinery planning are introduced, as well as the latest research and application of optimization algorithm in petrochemical industry, especially application of genetic algorithm. What's more, a simulation-based optimization framework is proposed. Later, this thesis's research objectives and innovative points are presented.In Chapter 2, the genetic algorithm's theory is systematically introduced, focusing on constraints handling and accelerating techniques. Experiences of parameters setting are listed in order to help the later solving application.In Chapter 3, nonlinear models of several important units are built and tank models are described in detail so that lots of inequality constraints are brought in. A two-scale constraints handling strategy is put forward as well as hybrid framework with the direct search method.In Chapter 4, modeling of uncertainties including demand and yield of refinery is implemented. After a case study of Monte Carlo sampling amount, yield model with Markov Chain is imported. Finally, multi-population parallel mode is described for improving solving performance.In Chapter 5, based on the fact that genetic algorithm has not been used on large-scale engineering applications, an integrated optimization platform is conducted. A factory model from real-live data is built which is linked with optimization subsystem centralized in genetic algorithm. Finally, the comparison of the platform and other software is made through software quality evaluation criteria.In Chapter 6, a summary of the research referred above is concluded and the prospect of future study is indicated in this dissertation.