Optimization of Sustainability in Process Design Based on Covariance Matrix Adaptation Evolution Strategy

Abstract

Preliminary design in chemical process furnishes economic feasibility through the calculation of both mass balance and energy balance, and it makes the process possible to produce a desired product under the given conditions. Through this design stage, the process possesses unchangeable characteristics, since the materials, reactions, unit configuration, and operating conditions are determined. Therefore, it becomes more important to design process considering sustainability. For this reason, this thesis proposes the solution procedure to integrate sustainability into traditional process economic optimization. The process modeling is conducted by the general-purpose sequential modular simulator employed for both convenience and reliability to analyze and design chemical process. However, using the sequential modular simulator is hard to obtain derivatives of process models for the deterministic optimization strategy, and it is also difficult to satisfy the process constraints and design specifications multidirectionally interacting with design variables. Therefore the covariance matrix adaptation evolution strategy (CMA-ES), which is a stochastic black-box optimization algorithm, is adopted to overcome these difficulties in this thesis. The CMA-ES is an improved methodology in terms of accuracy and reliability to find optimal solution in comparison with other stochastic sampling based algorithms, and it has several advantages that: 1) it has much fewer initial settings required by the user, 2) it can deal with the non-convex multimodal problem even though the problem contains discontinuous decision variables so that it solves the optimization problem which is too complicated to interpret mathematical model explicitly, and 3) it has an excellent ability to optimize the multi-variable problems. These advantages improve the performance of the solution procedure proposed in the thesis. The proposed solution methodology for optimizing the sustainability of the process is an iterative procedure which executes the nested loop consisting of the inner loop for the process model simulation and the outer loop for the economic and sustainability optimization. Thus, 1) the CMA-ES for finding an optimal solution with much fewer function evaluations and 2) the rejection method for avoiding the execution of simulator in case of infeasible individuals make the convergence of a nested loop faster as well as expand the searching domain economically. The effectiveness and usefulness of the proposed solution methodology in this thesis are verified by the application to practical process design problems. At first, the most profitable operating condition of an offshore oil and gas production process is determined with the consideration of the Reid vapor pressure specification and wastewater treatment cost as an environmental aspect. Second, regarding the inherently safer design of the offshore natural gas liquefaction process which involves the risk of explosion, the quantification of an inherent explosion consequence and integration of inherent risk into economic analysis make it possible to show the numerical relation between conflicted objectives and assist a decision-making to design inherently safer process. As a result of this thesis, the proposed methodology makes it possible to design the sustainable chemical processes. It is expected to be useful in designing the sustainable process since the sustainability factors in the process can be numerically monitored during preliminary process design stage using process simulator. Furthermore, it contributes to decrease various uncertainties during the process lifetime and minimize the risk and further expenses regarding the economic, environmental and safety aspects.