煤矿瓦斯突出预测的PSO-LSSVM模型

Prediction on Gas Emission Value Based on PSO- LSSVM

利用实际的瓦斯数据作为样本,由于瓦斯数据具有高维、非线性的特点,提出了用粒子群优化(PSO)对支持向量机(LSSVM)的惩罚参数(c)、核参数(σ)和损失函数(ε)参数进行优化,获得最优参数的组合。利用获得的最优参数建立粒子群优化支持向量机(PSO-LSSVM)瓦斯涌出量级别的预测模型。实验结果表明:PSO可以对LSSVM的参数进行很好的优化,通过参数优化的LSSVM对获得的瓦斯数据进行回归分析,建立的瓦斯预测模型是SVM模型预测精度的2倍,收敛时间比SVM缩短3倍,运算效率显著提高。

Because the high dimension and non-linear characteristics of the gas data,this article proposed a method that used the particle swarm optimization to optimize 3 parameters（punishment parameter（c）,nuclear parameters（σ） and loss function epsiIon （ε）） of support vector machine so as to obtain the optimal parameter combination.Using the optimal parameters obtained by particle swarm optimization support vector machine （PSO-LSSVM） to build gas emission level forecasting models.The experimental results show that the PSO has a very good optimization of LSSVM parameters.Through the parameter optimized LSSVM to make the regression analysis of the the obtained gas data.It shows that the accuracy of this gas prediction model is 2 times more and the convergence time is 3 times less than the SVM model,and the operational efficiency is improved significantly.