用于热力系统建模的基于粗糙集的模糊神经网络

模糊神经网络应用于热力系统建模,虽能取得较好的效果,但当模糊规则较多时,网络学习速度较慢。针对这个问题,对传统的模糊神经网络进行了改进。利用Kohonen自组织网络对数据信息进行聚类。然后利用粗糙集规则约减的方法,获取模糊神经网络最小规则,以提高模糊神经网络的学习速度。经过锅炉汽压回路模型的仿真实验结果表明:粗糙模糊神经网络学习速度较传统模糊神经网络有较大提高,同时网络误差有所降低。

Rough set-based fuzzy-neural network model design for thermodynamic systems

Fuzzy-neural networks can accurately identify thermodynamic systems, but the networks usually have a slow learning speed. This study presents an improved fuzzy-neural network method that applies Kohonen network clustering analysis to the data table, and then uses rough sets in the fuzzy-neural network to reduce the decision table size and to accelerate the approach to the minimal rules. In a stimulation of a boiler steam loop, the rough set-based fuzzy-neural network increased the learning speed and reduced the error. Therefore, the rough set-based fuzzy-neural network improves the performance of fuzzy-neural networks used for analyzing thermodynamic systems.