摘要
为解决多输入单输出数据集的建模预测问题,提出一种基于Mamdani型模糊系统和前向神经网络的模糊神经网络,实现了瓦斯涌出量的建模预测.首先由采样数据生成模糊规则,明确了前向神经网络的网络结构.在Mamdani型模糊系统中提取出了隐含层神经元激励函数,并据此确定了模糊前向神经网络的表达式.然后对BP学习算法进行了改进,得到了权值直接确定的矩阵式.最后在瓦斯涌出量预测中,利用主成分分析法选取了较为重要的3个因素.仿真实验表明模糊前向神经网络具有较高的建模和预测精度.
In order to solve the problem of multi-input single-output gas emission in modeling and predicting, the combination of fuzzy neural network was put forward on the basis of Mamdani fuzzy system and feed-forward neural network. Firstly, the fuzzy rules were generated by the sampled data, and then the network structure of the feed-forward neural network was determined. Secondly, the hidden layer neuron activation functions were extracted from the Mamdani .fuzzy system, and the expression of the fuzzy neural network was received. The BP learning algorithm had been improved, had been directly determined weights matrix. Finally, the three most important factors in the gas emission prediction were selected by the method of principal component analysis. The simulation results showed that the fuzzy feed-forward neural network had high rnndt^lin~ ~ncl nr^Aic, f;
出处
《安徽大学学报(自然科学版)》
CAS
北大核心
2013年第6期30-35,共6页
Journal of Anhui University(Natural Science Edition)
基金
中央高校基本科研业务费资助项目(3142013021)
华北科技学院高等教育科学研究资助课题(HKJYZD201213)
关键词
模糊前向神经网络
权值直接确定
瓦斯涌出量
预测
fuzzy feed-forward neural network
weights-direct-determination
gas emission quantity
prediction