基于改进型遗传算法的再热汽温神经网络模型构建及预测分析

Construction and Predict Analysis of a Neural Network Model of Reheated Vapor Temperature Based on Advanced Genetic Algorithm

在根据混沌理论非线性重构技术揭示火电机组再热汽温变化的混沌动力学特性的基础上,构建了基于改进型遗传算法(AGA)的再热汽温神经网络模型。该模型利用混沌特性处理输入样本及确定神经网络的网络结构,用神经网络映射混沌相空间相点演化的非线性关系,采用AGA对神经网络模型进行参数辨识。训练结果表明,该模型精度较高,收敛速度快,为生产实际过程中预测机组再热汽温提供了一种新的思路和方法。

Reconstruction in nonlinear chaotic theory is adopted to reveal the chaotic dynamics performance of reheated vapor temperature in power station. Chaotic performance is used to deal with input samples and determine structure of neural network, neural network mapping is used to describe nonlinear of point in reconstruction phase space, and parameter identification is done by advanced genetic algorithm（ AGA）. So a new predicting model of neural network based on AGA is made. By samples training, the model has higher precision and quick convergence speed, which is significant in predicting reheated vapor temperature.