Superstructure-free synthesis and optimization of thermal power plants

Abstract

The optimal synthesis and design of thermal power plants is best addressed using mathematical optimization techniques. However, conventional optimization methods require the user to manually define a priori the potential solution space through the modeling of a superstructure, which is a time-consuming, complex, and error-prone task. More importantly, the final superstructure model can not guarantee to contain all good alternatives (in particular, the optimal solution), while it might consider a huge number of meaningless or even infeasible alternatives. To circumvent the use of superstructures, recently, a generic superstructure-free optimization approach has been proposed by the authors to synthesize distributed energy supply systems. The approach employs hybrid optimization integrating evolutionary and deterministic optimization to enable simultaneous alternatives generation and optimization. In this paper, the approach is extended by a new mutation rule to enable the synthesis of thermal power plants. The features of the extended superstructure-free approach are illustrated by a case study. The optimization is initialized with the simplest plant cycle consisting of one turbine, one pump, one steam generator, and one condenser. The superstructure-free approach automatically identifies highly efficient and complex structures featuring multi-stage reheating and feedwater preheating.