基于CSO-RVM的瓦斯涌出量预测模型研究

为了实时监测和精准预测煤矿回采工作面绝对瓦斯涌出量,提出猫群算法(CSO)优化相关支持向量机(RVM)的绝对瓦斯涌出量预测方法.相关向量机的组合核函数可实现多特征空间的信息融合,为有限样本、高维数瓦斯涌出量预测建模问题提供一种行之有效的方法.并用CSO算法对RVM瓦斯涌出量预测模型的核函数权重p和高斯核参数σ快速寻优.利用矿井无线传感器网络检测到的各项历史数据试验.结果表明,相比BP、SVM算法,该耦合模型有效提高了预测精度,具有更好的泛化能力,为矿井瓦斯预测提供理论支持.

Prediction model research of gas emission based on CSO-RVM

In order to real-timely supervise and accurately predicate the absolute gas emission in mines,the absolute gas emission prediction method which uses Cat Swarm Optimization(CSO)to optimize Relevance Vector Machine(RVM) is proposed. Multi-kernel learning function of RVM can help to realize information fusion in multi-feature space,thus provides a practical method for gas emission prediction model which with limited samples and high dimension. Kernel function weight p of RVM gas emission prediction model and Gauss kernel parameterσcan fast get the best value by CSO algorithm. The data from mine wireless sensor networks are used for experiments. The experimental result shows that the coupling model improve the prediction precision effectively and it has better generalization ability to provide theoretical support for mine gas prediction,when compared with BP algorithm or SVM algorithm.