一种基于神经网络的航改燃气轮机转速控制器

A Neural Network-Based Speed Controller of an Aero-Derivative Gas Turbine

关键词:  航改燃气轮机  转速控制  PI控制器  反正切神经网络  集成控制  仿真测试

Key words:  Aero-derivative gas turbine  Speed control  Proportional-integral controller  Arctangent neural network  Integrated control  Simulation test

摘要:

为了提升某型航改燃气轮机的转速控制性能，提出一种双输入反正切神经网络控制器，然后基于该航改燃气轮机的PI控制器以及该反正切神经网络控制器，提出一种新型的集成控制器。采用仿真计算的方法，通过跟踪测试、抗干扰测试和鲁棒性测试，对比此三种控制器以及四种常规控制器的性能差异。结果表明，在三次跟踪测试中，集成控制器具有最优的综合跟踪性能；给燃料量加入5%的干扰后，集成控制器的抗干扰能力等于反正切神经网络控制器，但高于另外五种控制器；当工作环境温度增加以及压气机性能退化引起不确定性时，七种控制器均能正常实施控制，且仍以集成控制器的效果最优，其鲁棒性最强。新型集成控制器具有最佳的综合转速控制性能，能够保证该型航改燃气轮机安全和高效的运行。

Abstract:

In order to improve the speed control performance of a certain aero-derivative gas turbine， an arctangent neural network controller with dual inputs was proposed. Then， a novel integrated controller was proposed based on the proportional-integral（PI） controller of this aero-derivative gas turbine and this arctangent neural network controller. By adopting the method of simulation calculation， tracking test， disturbance-rejection test and robustness test were carried out to compare the performance difference of these three controllers and four common controllers. The results show that during the three tracking tests， this integrated controller has the optimal comprehensive tracking performance. After 5% disturbance is added to fuel flow rate， the disturbance-rejection ability of this integrated controller is equal to that of arctangent neural network controller， but higher than that of another five controllers. When increasing operational ambient temperature and the deterioration of compressor performance cause the uncertainties， seven controllers all play the normal control role. In addition， the effect of this integrated controller is still the best， thus the robustness of it is the strongest. This novel integrated controller has the optimal comprehensive speed control performance， and is able to guarantee the safe and high-efficient operation of this aero-derivative gas turbine.