煤矿突水水源判别的PCA-PSO-ELM模型

Discriminant model of coal mine water inrush based on PCA-PSO-ELM

在矿井水害防治工作中,对突水水源类型的快速和准确识别尤为重要。选取焦作矿区中36组不同含水层的水样数据,将Ca^(2+)、Mg^(2+)、K^(+)+Na^(+)、HCO_(3)^(-)、Cl^(-)和SO_(4)^(-2)这6种因子作为评价指标,使用主成分分析(principal component analysis,PCA)算法进行降维,消除叠加信息对预测结果的影响,利用粒子群优化(particle swarm optimization,PSO)算法优化极限学习机(extreme learning machine,ELM)神经网络的初始权值和阈值,克服ELM神经网络输入层权重和隐含层偏置具有随机性且隐含层很难确定的缺点,最终建立基于主成分分析-粒子群优化算法-极限学习机(PCA-PAO-ELM)的突水水源判别模型。对比ELM、BP神经网络模型可以看到,经过PCA降维和PSO改进参数的ELM神经网络模型解决了传统模型易陷入局部极小值点和学习过程收敛速度慢的问题,减小了水源识别的误差,提高了模型的泛化性,使预测结果更加可靠,为快速识别突水水源提供了新的思路。

In order to prevent and control mine water disaster,it is particularly important to accurately identify the type of water inrush source.Water sample data of 36 groups of different aquifers in Jiaozuo mining area were taken.Six factors including Ca^(2+),Mg^(2+),K^(+)+Na^(+),HCO^(-)_(3),Cl^(-)and SO^(-2)_(4)were taken as evaluation indexes.Principal component analysis(PCA)method was used to reduce dimension and eliminate the influence of superposition information.In order to overcome the disadvantage of the randomness of input layer weight and hidden layer bias of extreme learning machine(ELM)neural network and the difficulty in determining the hidden layer,this paper uses particle swarm optimization(PSO)algorithm to optimize the initial weight and threshold value of ELM,and finally establishes the PCA-PSO-ELM water source discrimination model.The prediction results of the ELM neural network model with improve parameters after PSO are compared with those of the traditional ELM and BP neural network models.it can be seen that PCA dimensionality reduction can eliminate the influence of redundant information,PSO can solve the problem that elm is easy to fall into local optimization,and improve the convergence ability and global search ability of the algorithm,Thus,the elm neural network model optimized by particle swarm optimization has higher generalization,more reliability and less water source identification error,which provides a new idea for the rapid identification of water inburst sources.