Real-Time Variable Geometry Triaxial Gas Turbine Model for Hardware-in-the-Loop Simulation Experiments

Abstract

This paper proposes a hybrid method (HMRC) comprised of a radial basis function (RBF) neural net algorithm and component-level modeling method (CMM) as a real-time simulation model for triaxial gas turbines with variable power turbine guide vanes in matlab/simulink. The sample size is decreased substantially after analyzing the relationship between high and low pressure shaft rotational speeds under dynamic working conditions, which reduces the computational burden of the simulation. The effects of the power turbine rotational speed on overall performance are also properly accounted for in the model. The RBF neural net algorithm and CMM are used to simulate the gas generator and power turbine working conditions, respectively, in the HMRC. The reliability and accuracy of both the traditional single CMM model (SCMM) and HMRC model are verified using gas turbine experiment data. The simulation models serve as a controlled object to replace the real gas turbine in a hardware-in-the-loop simulation experiment. The HMRC model shows better real-time performance than the traditional SCMM model, suggesting that it can be readily applied to hardware-in-the-loop simulation experiments.