基于遗传算法与人工神经网络的加热炉建模方法研究

发布时间：2018-01-05 04:09

本文关键词：基于遗传算法与人工神经网络的加热炉建模方法研究　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：加热炉是一个特性复杂的工业生产对象,对于加热炉的建模,传统的建模方法往往不够精确,投运效果不尽如人意。随着智能技术的蓬勃发展,以BP神经网络为代表的神经网络建模方法在工业过程中的应用日渐广泛。遗传算法作为一种进化计算方法,具有的全局搜索能力能够较好地克服BP神经网络容易陷入局部极小值的缺点。因此,研究将BP神经网络与遗传算法相结合的方法十分必要。本文提出了一种基于历史数据的加热炉神经网络建模方法。该方法首先确定BP神经网络结构得到网络模型的参数个数。然后通过遗传算法对编码后的个体反复地选择、交叉、变异,最终获得的最优个体包含BP神经网络权值和阈值的最佳参数组合,将其解码后作为BP神经网络的初始值。最后通过BP算法训练得到加热炉对象的神经网络模型。本文所做的主要工作与贡献有以下几点:(1)针对加热炉对象的复杂特性,本文从对象的输入输出特性出发,研究了利用神经网络对加热炉炉温建模的有效性,通过对比有效性指标,解决了神经网络结构难以选择的问题。(2)本文提出了将遗传算法应用于加热炉神经网络建模中,通过遗传算法的大规模全局搜索能力优化BP神经网络模型的权值和阈值,避免了网络陷入局部极小值的缺点,加快了网络的收敛速度。基于该方法获得的模型学习能力更强,预测效果更好。(3)为了加快遗传算法的收敛速度,提高运算效率,本文基于加热炉的运行数据,对遗传算法进行参数自适应调整。设计并实现了变异概率自适应的遗传算法,加快了神经网络模型参数的收敛速度。(4)本文以唐山某钢铁厂的加热炉为研究对象,选取该对象的样本数据来设计加热炉神经网络模型,通过对比验证了该方法的可行性。结果表明应用该方法进行加热炉炉温建模是行之有效的。
[Abstract]:Reheating furnace is a complex industrial production object. The traditional modeling method is often not accurate enough and the operation effect is not satisfactory. With the rapid development of intelligent technology. BP neural network modeling method, represented by BP neural network, is widely used in industrial process. Genetic algorithm (GA) is an evolutionary computing method. The global search ability can overcome the shortcoming that BP neural network is easy to fall into local minima. It is necessary to study the method of combining BP neural network with genetic algorithm. In this paper, a neural network modeling method for heating furnace based on historical data is proposed. This method first determines the BP neural network structure to get the network. The number of parameters of the model. Then the genetic algorithm is used to select the coded individuals repeatedly. Crossover, mutation, and the final optimal individuals include BP neural network weights and threshold values of the best combination of parameters. After decoding it as the initial value of BP neural network, the neural network model of heating furnace object is trained by BP algorithm. The main work and contribution of this paper are as follows: 1). Aiming at the complex characteristics of heating furnace object. Based on the input and output characteristics of the object, this paper studies the effectiveness of modeling the furnace temperature with neural network, and compares the validity index. It solves the problem that the neural network structure is difficult to select. (2) in this paper, the genetic algorithm is applied to the neural network modeling of heating furnace. The weight and threshold of BP neural network model are optimized by the large scale global search ability of genetic algorithm, which avoids the shortcoming that the network falls into local minima. Based on this method, the model learning ability is stronger, the prediction effect is better.) in order to speed up the convergence speed of genetic algorithm, improve the operation efficiency. Based on the operation data of heating furnace, this paper adaptively adjusts the parameters of genetic algorithm, and designs and implements a genetic algorithm with adaptive mutation probability. This paper takes the heating furnace of a steel plant in Tangshan as the research object, and selects the sample data of the object to design the neural network model of the heating furnace. The feasibility of the method is verified by comparison and the results show that the method is effective in modeling furnace temperature.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG307;TP18

【参考文献】