基于GASA-SVR的矿井瓦斯涌出量预测研究

Research on Prediction of Mine Gas Emission Quantity Based on GASA-SVR

针对煤矿工作面瓦斯涌出量的多影响因素、非线性、时变性和不确定性等特点,提出了遗传模拟退火算法(GASA)与回归型支持向量机(SVR)的耦合算法(GASA-SVR)用于瓦斯涌出量预测。利用煤层瓦斯含量、深度、厚度、倾角等12个参数作为主要影响因素,经过归一化处理后作为回归型支持向量机训练和测试样本。采用遗传模拟退火算法寻找最优的惩罚参数和核函数参数,同时引入自适应交叉和变异概念,建立瓦斯涌出量的非线性拟合模型,并利用矿井实测历史数据进行试验,结果表明该预测模型比传统的神经网络模型具有更理想的精度和稳定性,可为煤矿瓦斯爆炸的防治提供可靠的理论依据。

Genetic simulated annealing algorithm（GASA）and regression support vector machine（SVM）coupling algorithm（GASA-SVR）were presented to predict gas emission in the view of the coal mine characteristics such as complicated,time varying and nonlinear etc.12 parameters,such as gas content,depth,thickness and dip angle of coal seam,were used as the main influencing factors,the training and test samples were trained and tested as regression support vector machine.Genetic simulated annealing algorithm was used to find the optimal penalty parameter and kernel parameter.Simultaneously,the concept of adaptive crossover and mutation is introduced,the nonlinear fitting model of gas emission is established,and the experiment was carried out by using the measured historical data of the mine.The results show that the prediction model has better accuracy and stability than the traditional neural network model,which can provide a reliable theoretical basis for the prevention and control of coal mine gas explosion.