In the process of a gas turbine performance simulation based on component characteristics, the accuracy description of components characteristics is a significant element which affects the accuracy of the results, especially for part-load and startup performance calculation.
For the traditional modeling method of components, the discrete characteristic maps was integrated into the thermodynamic model in the form of arrays, which present the relationships among IGV opening angle γ, speed, pressure ratio and flow/efficiency of compressor, as well as the relationships among speed, pressure ratio and flow/efficiency of turbine. It’s obvious that fine discrete grids are required in this traditional interpolating method to obtain enough data to reflect the original information accurately. However, the characteristic maps obtained are always sparse, which lead to considerable interpolation error.
With regard to poor accuracy of traditional component characteristics method, we integrate partial least square (PLS) method into reconstructing the components characteristic curves. It doesn’t need the discretization of component characteristic maps, and the reconstruction equation can be solved directly in the simulation process, which avoids the interpolation errors in the traditional processing methods of component characteristics.
The traditional method, BP artificial neural network, PLS of two variables and ternary PLS of three variables was implemented in performance simulation of the same gas turbine. The simulation results of components characteristics and gas turbine thermodynamic performance with different methods were compared with the known performance data of the gas turbine. And the calculation accuracies and their effects on overall performance were analyzed. The implement of PLS in component characteristics simulation could be useful for improving the accuracy of the gas turbine thermodynamic model, and could benefit for gas turbine power plant upgrades and technology improvement because thermodynamic mode is the base of on-line monitoring and fault diagnosis system.