动力涡轮导叶可调的三轴式船用燃气轮机动态实时仿真

Dynamic Real-Time Simulation of Triaxial Marine Gas Turbine with Variable Power Turbine Guide Vane

关键词:  导叶可调  船用燃气轮机  实时仿真  RBF神经网络  部件建模法

Key words:  Variable guide vane  Marine gas turbine  Real-time simulation  RBF neural net  Component modeling method

摘要:

鉴于传统的燃气轮机建模方法用于动力涡轮导叶可调的三轴式船用燃气轮机动态实时仿真时具有一定局限性，提出了一种新型的基于径向基函数(Radical basis function，RBF)的神经网络与部件法(Component modeling method，CMM)的复合建模方法(HMRC)。该方法针对该型燃气轮机的结构和总体性能特点，合理地选取了样本点以减小样本规模。通过神经网络与部件法分别计算燃气发生器与动力涡轮相关参数，采用MATLAB/SIMULINK仿真软件建立实时仿真模型。与传统的部件法建立起来的仿真模型对比，新方法计算结果与部件法吻合度高，误差低于1%，同时模型的实时性得到了大幅提升，计算速度约为部件法的7倍。

Abstract:

There are some limitations in using traditional gas turbine modeling approaches for dynamic real-time simulation of triaxial marine gas turbine with variable power turbine guide vane. Therefore， Hybrid Method (HMRC) of Radial basis function (RBF) neural net algorithm and Component modeling method (CMM) is developed. After analyzing the characteristics of the gas turbine structure and overall performance， sample points are selected reasonably to minimize sample size. Using neural net algorithm and component modeling method to calculate gas generator and power turbine relevant parameters respectively， the real-time simulation model is built by MATLAB/SIMULINK. Through the comparison between new model and CMM model， computational results of the new model are very close to those of CMM model with the error less than 1%. Meanwhile， the new model real-time performance is improved significantly as the new model is about 7 times CMM model computation speed.