基于改进极端学习机的混沌时间序列瓦斯涌出量预测

Prediction of Chaotic Time Series of Gas Emission Based on Improved Extreme Learning Machine

为更准确地预测瓦斯涌出量,预防瓦斯灾害,有必要建立和应用基于改进极端学习机(IELM)的混沌时间序列预测模型。首先,对瓦斯涌出量监测数据构成的多变量时间序列进行相空间重构,采用互信息法与虚假邻点法得到每一变量的延迟时间和最佳嵌入维数;然后,通过最小二乘方法和误差反馈原理计算出最优的网络输入层到隐含层的学习参数,对极端学习机(ELM)进行改进;最后,借助IELM建立瓦斯混沌时间序列的预测模型。通过仿真试验,运用该预测模型预测的最大相对误差为3.290 2%,最小相对误差为0.898 2%,平均相对误差为1.952 8%。

It was held by the authors that in order to predict the amount of gas emitted more accurately, and then to prevent the occurrence of gas disasters, a chaotic time series prediction model based on the improved extreme learning machine（IELM） should be built and used. For this end, the multivariate time series consisted of the gas emission monitoring data were reconstructed using phase space. The delay time and the best embedding dimension for each variable were obtained using the mutual information method and false neighbor. The optimal learning parameters from network input layer to hidden layer were calcu- lated using the least-square method and error feedback principle. The ELM was so improved, lastly, a gas chaotic time series prediction model was built with the help of IELM. Through the simulation experiment, the prediction model has predicted that the maximum relative error is 3.290 2 %, the minimum relative error is 0. 898 2 % and the average relative error is 1. 952 8 %.