基于加权极端学习机的瓦斯涌出量预测模型

Prediction of Gas Emission Based on Weighted Extreme Learning Machine

为了能够更加准确地预测瓦斯涌出量，提出一种多元混沌时间序列的加权极端学习机瓦斯涌出量预测模型。首先对瓦斯涌出量监测数据构成的多元时间序列进行相空间重构，采用信息熵方法选取最佳延迟时间和嵌入维数：然后根据相空间中输入数据对预测误差的影响施加不同的权重，并结合核极端学习机预测模型构造出加权极端学习机模型。通过仿真试验表明，提出的预测模型行之有效，与同类其他模型相比，具有更高的预测精度和更好的稳定性。

In order to predict the amount of gas emitted more accurately, a multivariate chaotic time series prediction model based on weighted extreme learning machine is proposed in this paper. Firstly, the multivariate time series consisted of the gas emission monitoring data are reconstructed using the phase space. The delay time and the best embedding dimension are obtained using the information entropy method. Then, the instances are weighted according to their influence on the prediction errors and the model of weighted extreme learning machine is constructed based on kernel extreme leaming machine. Finally, a chaotic time series prediction model of gas emission is built with the help of WELM. It is demonstrated by the simulation experiment that the proposed model works effectively. And compared with other similar prediction methods, the proposed model performs better in terms of prediction accuracy and robustness.