摘要
燃煤电厂锅炉结渣问题是一个复杂的物理化学过程,影响着锅炉运行的安全性、经济性,通过建立专家系统知识库和推理机制,模拟人类专家的思维过程,可以完成对锅炉结渣程度的预测。首先采用模糊方法对样本数据进行模糊化处理以符合建模数据的模糊特性,继而采用误差反向后传(Back Propagation,BP)神经网络的建模方法,把训练数据集隐含的专家知识变换为神经网络内部的表达形式,构建出由神经网络的权值、阈值组成的知识库,并使用遗传算法(Genetic Algorithm,GA)对网络权值、阈值进行优化;其次在由历史数据获得的知识库的基础上,建立输入模糊值与输出模糊值之间的模糊神经网络(Fuzzy Neural Network,FNN)推理结构。为体现文中提出的GA-FNN方法的优越性,将GA-FNN法同不考虑输入数据模糊性的GA-BP法以及未对知识库中的权值、阈值进行优化的FNN法进行了对比,结果表明基于GA-FNN的专家系统可以对锅炉结渣程度进行更准确地预测。
In operation of coal-fired boilers in power plants,slagging is a complex physical and chemical process that affects the safety and economy of the boiler.The degree of boiler slagging can be predicted through the establishment of expert system knowledge base and inference mechanism that can simulate the thought process of a human expert.We first use fuzzy method to deal with the sample data whose fuzzy nature can be presented,and then use Back Propagation(BP)modeling method to transform the implicit knowledge of the training data for neural network internal representations,which can build knowledge base composed of neural network weights and threshold by using a Genetic Algorithm(GA) to optimize.On the basis of the knowledge base obtained from historical data,we establish the Fuzzy Neural Network(FNN)inference structure between the input fuzzy value and the output fuzzy value.In order to manifest the superiority of the proposed GA-FNN,we compare the GA-FNN method with the GA-BP method which does not consider the fuzziness of input data and the FNN method without optimization of weights and threshold,the results show that expert system based on GA-FNN can accurately predict the degree of boiler slagging.
出处
《电站系统工程》
北大核心
2012年第6期1-4,9,共5页
Power System Engineering
基金
国家自然科学基金项目(61174114
60574047)
国家高技术研究发展计划项目(2007AA04Z168)
教育部博士点基金项目(20050335018)