Gradient-free methods applied to optimisation of advanced ultra-supercritical power plant

Thermo-economic analysis and optimisation of thermal systems have become a key solution in providing a better system configuration on the design stage and in modification of already-working installation. Nowadays, there are many commercial software packages allowing one to perform 0-dimensional (0D) analysis of any power plant system. Some of these programs are equipped with their own optimisation tools; however, as is often the case, most of them are not able to carry out a multi-variable optimisation with a large number of decision variables. This paper presents application of the integrated package containing the IPSEpro and MATLAB software for numerical optimisation of advanced ultra-supercritical steam power plant. Searching for new technology makes it necessary to analyse the structure of the thermal cycle. This can be obtained by combining the numerical modelling with the multivariable optimisation of such an object. For that purpose a new concept of 900 MW advanced ultra-supercritical power unit was examined. To increase the efficiency and flexibility of calculations three different non-gradient methods were introduced into the code, i.e., Nelder-Mead, Hooke-Jeeves and Rosenbrock. It was shown that all of the three considered methods can find their usefulness in the optimisation process of power plant thermal cycle giving the rise in the efficiency by about 0.34 percentage points. The additional significant advantage of the Rosenbrock method is the lowest cost of computation, almost independent of the number of decision variables considered. It was shown that the gained efficiency improvement was accompanied with serious intervention into the thermal cycle configuration (in particular into operating pressures) which can lead to necessity of modifying the major power plant components.