基于ELM的矿井突水水源识别方法

Identification Method of Mine Inrush Water Source Based on ELM

针对矿井突水事故的预测问题,提出一种基于极限学习机(Extreme Learning Machine,ELM)的矿井突水水源识别新方法。该方法是一种单隐含层前馈神经网络学习算法,在训练过程中无需调整初始连接权值和阈值,只需要设置隐含层神经元个数即可获得最优解。以梧桐庄煤矿水质为例,通过MATLAB仿真证实,该方法不仅克服了常规BP神经网络受初始权值和阈值影响的缺陷,而且识别精度更高;在突水预测方面有很好的应用前景。

In view of the prediction of mine water inrush accident,a new method for identifying inrush water source based on Extreme Learning Machine（ ELM） was proposed. This method was a single hidden layer feedforward neural network learning algorithm. There is no need to adjust the initial connection weights and thresholds during the training process. The optimal solution can be obtained only by setting the number of neurons in the hidden layer. Taking the water quality of Wutongzhuang Coal Mine as an example,it was confirmed by MATLAB simulation. This method not only overcame the defect that the conventional BP neural network was affected by the initial weight and threshold,but also has a higher recognition precision. It has a good application prospect in the prediction of water inrush.