改进的遗传神经网络在灰渣粘度预测中的研究及应用

Research and Application of An Improved Genetic Neural Network in Slag Viscosity Prediction

提出一种基于改进遗传神经网络的灰渣粘度预测方法。通过节点相关性评价,使得低于某阈值的网络内部节点在交叉操作时被排除,从而降低节点冗余;将交叉/变异概率与种群个体适应度相联系,提出一种自适应交叉/变异概率,将其用于遗传操作,使得个体多样性较好。仿真实验表明,用提出的算法优化的神经网络在一定程度上可以避免"种群早熟",保持种群多样性,提高了学习效率。

This paper presents an ash viscosity prediction method based on an improved genetic neural network.Nodes whose node-related evaluation is below a certain threshold within the network are excluded in the cross-operation by the node-related evaluation,which reduces the node redundancy.An adaptive crossover-mutation probability used in GA improves the population diversity,which is presented in this paper.Simulation results show that the neural network optimized with the proposed algorithm can avoid prematurity,keep the population diversity,and accelerate learning efficiency.