基于网格优化双层随机森林的采空区煤氧化升温预测研究

Prediction of coal oxidation temperature rise in goaf based on grid optimizationdouble-layer random forest

为了对采空区煤氧化升温的温度进行预测,在内蒙古某煤矿16402综放工作面进行长期的采空区气体和温度观测实验,采集到准确的采空区煤氧化升温过程中气体及温度数据,提出1种基于网格优化双层随机森林(WG-DRF)的采空区煤氧化升温预测方法,用该方法构建预测模型并与传统随机森林、BP神经网络和支持向量回归模型的预测结果进行对比。研究结果表明:WG-DRF模型预测的平均绝对误差MAE,均方误差MSE,决定系数R~2分别为1.725,6.158,0.903,优于其他模型。通过更换数据集对WG-DRF方法进行测试,验证双层随机森林模型具有较强的泛化性。研究结果可为采空区煤氧化升温的温度预测提供参考。

In order to predict the temperature of coal oxidation temperature rise in goaf,a long-term observation experiment of goaf gas and temperature was carried out on the 16402 fully mechanized caving face of a coal mine in Inner Mongolia to collect accurate gas and temperature data during the process of coal oxidation heating in goaf.A method for predicting the coal oxidation temperature rise in goaf based on the grid optimization double-layer random forest(WG-DRF)was proposed.The prediction model was constructed by this method and compared with the prediction results of traditional random forest,BP neural network and support vector regression model.The results show that the mean absolute error MAE,mean square error MSE and coefficient of determination R 2 of WG-DRF model are 1.725,6.158 and 0.903,respectively,which are better than the other models.The WG-DRF method is tested by changing the data set,and it verified that the double-layer random forest model has strong generalization.The research results can provide reference for the temperature prediction of coal oxidation temperature rise in goaf.