基于MEA的粗糙集神经网络研究及应用

Research and Application of Rough Set-BP Neural Network Based on MEA

将思维进化算法、粗糙集和神经网络相结合,提出一种基于MEA的粗糙集神经网络,用于变压器故障诊断。此模型采用思维进化算法全局寻优的特点,搜索粗糙集属性约简离散断点的位置以及神经网络的连接权值和阈值,避免了常规粗糙集属性约简时复杂的手工试凑以及BP神经网络收敛速度慢、精度不高等缺点,有利于更快地收敛于全局最优解,提高系统的诊断速度和准确率。仿真结果表明了方法的有效性。

The mind evolutionary algorithm is combined, the rough set and the neural network, and a rough setneural network based on MEA is proposed applying in transformer fault diagnosis. This model uses global optimization characteristics of the mind evolutionary algorithm to search rough set discrete breakpoints and neural network connection weights and thresholds, it avoids the conventional rough set complex handwork reduction and slow convergence and low precision of BP neural network, and benefits to find the global optimal solution quickly and improves the diagnostic speed and accuracy. Simulation experiment verifies the validity of this method.