Research On Multi-objective Optimization Of Gas Turbine Based On Surrogate Model

As the main propulsion and power generation device of marine power system,gas turbine has the characteristics of multi-input,multi-output,strong coupling and nonlinearity.Based on the physical model of gas turbine,the multi-objective optimization of its operation economy,emission and power performance is realized in different environments and working conditions.The solving process is complex,the amount of calculation is large,and the real-time performance is poor.In order to improve the optimization efficiency,this paper establishes a gas turbine surrogate model to simplify the solution process.Based on the effective surrogate model of gas turbine,this paper studies the improved scheme of static and dynamic multi-objective optimization algorithm,and obtains the optimal operating points of gas turbine.The specific work is as follows:(1)Based on the basic principle of three axis gas turbine,Simulink simulation models are constructed by modular modeling method.According to the input-output relationship of the system,8 decision variables and 3 target variables are determined.The accuracy of the simulation model is verified by comparing the simulation values and the real values of the steady-state working condition of gas turbine under rated load.(2)According to the spatial distribution and filling index of gas turbine steady-state process sampling points,the appropriate design of experiment is selected.After comparing the four main surrogate modeling methods,RBF model is selected to build gas turbine surrogate model.Considering the accuracy and stability of the model,an ensemble of RBF surrogate model based on different combinations of radial basis functions is proposed.On this basis,the surrogate model is constructed to comprehensively reflect the economy,emission and power performance of gas turbine.(3)In the process of gas turbine static multi-objective optimization,NSGA-II algorithm is improved from crossover operator,mutation operator,crowding distance calculation method and algorithm termination condition,which verifies the advantages of the improved NSGA-II algorithm.On this basis,a method of generating robust multi-objective optimization solutions based on Pareto frontier is proposed.(4)The dynamic multi-objective optimization problem of gas turbine is transformed into a dynamic multi-objective optimization problem based on linear transformation of environmental variables,and a new environmental response mechanism is proposed based on NSGA-II algorithm.By setting the environment change detection mechanism,the original initialization and mutation strategies are integrated,and the adaptive initialization and mutation ratio are introduced.Numerical simulation shows that the proposed NSGA-II-C algorithm is more suitable for gas turbine dynamic multi-objective optimization problems.