基于神经网络的汽轮发电机汽门开度控制器

A STUDY ON THE VALVING CONTROLLER OF THE NEURAL NETWORK BASED TURBINE

汽轮发电机汽门开度调节过程是一个连续非线性动态过程。本文将之离散化并用多层ＢＰ网络逼近这一过程，构造了三种结构不同的神经网络汽门开度控制器。数字仿真测试证实了设计方案的可行性。基于对仿真结果的对比分析，文中讨论了神经网络对电力系统动态特性的抽取能力，并分析了样本选择与离线训练精度对控制器性能的影响。

The regulation of the turbine valve is a continuous nonlinear dynamic process.This regulation process is decentralized and achieved by the multi layer BP type neural network.Three neural network based turbine valving controllers (NNGOV) are constructed and discussed in the paper.Results from the digital simulation tests have proved the feasibility of the design.The effect of the neural network on the power system with dynamic characteristics is analyzed on the basis of the test results.Problems in the choice of training samples,the design of the neural network structure and the off line training are also discussed.