One of the most troublesome problems in the development of a component-based engine model is the compressor modeling because of the strong dependence of its performance on rotational speed and the treatment of compressor characteristic curves plays a significant role in modeling and simulation of gas turbine for the analysis of its off-design performance and thus it is crucial to describe compressor map exactly.
Usually part of rotational speed characteristic curves of compressor including on-design operating point are known during actual modeling and simulation, and reasonable interpolation and extrapolation have to be done in order to make good use of flow characteristics and efficiency characteristics under more constant rotational speed lines during off-design performance simulation.
Due to that the accuracy of traditional approaches for component characteristic treatment is not satisfactory, and the performance of interpolation and extrapolation of artificial neural network is poor, a linear multiple regression method, i.e., partial least-squares regression method was proposed in this paper.
Partial least-squares regression modeling method was used to reproduce the compressor maps. Different polynomial functions of various powers were used to obtain respectively expressions of compressor maps. Fitting accuracy and performance of interpolation and extrapolation of this method were analyzed, and the simulating experimental results show that partial least-squares regression method can ensure good fitting accuracy and good performance of interpolation and extrapolation for compressor thermodynamic modeling with both maximum RMS errors less than 0.5%.
It can be expected that the application of partial least-squares regression modeling has a certain reference value to improve the solution accuracy for thermodynamic model of industrial gas turbine.