粗糙集-改进神经网络落煤瓦斯涌出量预测  

Prediction for gas emission quantity of dropped coal based on rough set-improved neural network

作　　者：董晓雷[1,2] 贾进章[1,2] 樊程程[1] 赫祥林 

机构地区：[1]辽宁工程技术大学安全科学与工程学院,辽宁阜新123000 [2]矿山热动力灾害与防治教育部重点实验室,辽宁阜新123000

出　　处：《辽宁工程技术大学学报（自然科学版）》2014年第5期587-591,共5页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(51374121);辽宁省高等学校杰出青年学者成长计划基金资助项目(LJQ2011028)

摘　　要：为对井下落煤瓦斯涌出量进行预测,采用粗糙集与改进神经网络相结合的方法,在样本数据的筛选上吸取粗糙集数据约简的优点,使选择的数据样本简洁且更具代表性;充分利用BP神经网络的非线性拟合能力,将遗传算法与其相结合,避免BP网络陷入局部最优.利用编写的程序确定隐含层节点数,相比以往经验公式取值更具优势.最后在任家庄煤矿成功应用.研究结果表明:利用粗糙集与改进神经网络相结合模型进行预测,结果准确可靠,克服了以往BP模型的不足.该模型对井下落煤瓦斯涌出量预测具有一定参考价值.In order to predict the gas emission quantity of dropped coal, this paper applied the method of rough sets combined with the improved neural network, and selected the sample data by taking the advantages of rough sets data reduction, which is concise and more representative. By making full use of the nonlinear fitting capability of BP neural network, this study combined it with genetic algorithm, which avoided the local optimum of BP network. At the same time, the determination of hidden layer nodes was achieved by using the developed program, which is better than the past experience formula. At last, this new method was successfully used in coal mine of Renjiazhuang. The results show that the prediction of using the rough sets combined with improved neural network is accurate and reliable, overcoming the disadvantages of BP model in the past. This prediction model has certain reference value for gas emission quantity of dropped coal.

关 键 词：粗糙集 BP神经网络 落煤瓦斯 遗传算法 数据约简 涌出量 解析强度 样本数据 

分 类 号：TD712[矿业工程—矿井通风与安全]