基于WT与GALSSVM的瓦斯涌出量预测

Gas Emission Forecast Based on Wavelet Transform and Genetic-Least Square Support Vector Machine

为了准确预测工作面瓦斯涌出量,加强煤矿安全生产,基于小波变换(WT)和进化最小二乘支持向量机(GALSSVM),建立了瓦斯涌出量的新型预测模型.首先,通过小波分解将瓦斯涌出量时间序列分解成具有不同频率特征的信号;然后利用互信息法和伪近邻法得到的时间延迟和嵌入维数对各信号进行相空间重构;之后根据各个相空间的特点建立相应的GALSSVM预测模型;最后把各信号的预测结果进行小波重构,作为最终的瓦斯涌出量的预测结果.以晋城市成庄矿2315综放工作面瓦斯涌出量为例,进行了预测研究.实例表明,该方法具有很高的预测精度和较强的泛化能力.

In order to forecast the gas emission volume from working face and ensure the coalmine safety production,a novel model based on wavelet transform（WT） and genetic algorithm-least square support vector machine（GALSSVM） for gas emission forcast is presented.Firstly,gas emission time series is decomposed into different frequency signals through the wavelet transform;Secondly,phase space of each signals is reconstructed,and time delay and embeding dimension are determined by mutual information method and false nearest neighbor mehod respectively.Then,the respective forcasting model of GALSSVM is constructed according to different characteristics of each phase space.Lastly,the predicted results of the signals are reconstructed to be used as the final prediction result of gas emission.Gas emission from working face No.2315 of Cheng zhuang mine in Jing cheng city was predicted by using this model.The calculation result shows that this model has a higher forecasting precision and greater generality ability.