摘要
为了对煤矿井下瓦斯涌出量进行预测,采用粗糙集与改进极限学习机相结合的方法,在样本数据的筛选上吸取粗糙集数据约简的优点,充分利用极限学习机训练速度快、具有良好泛化性能的特点,并结合遗传算法选择最优的输入权值矩阵和隐含层偏差,避免随机产生所造成的误差。利用编写程序确定隐含层神经元个数,比依靠经验更为准确。在实际应用中选取煤层瓦斯含量、煤层埋藏深度、煤层厚度、煤层间距、工作面日产量五个因素作为预测的影响参数。研究结果表明:该预测模型预测的最大相对误差为5.6871%,最小相对误差为0,平均相对误差为2.5827%,相比改进前的预测模型具有更强的泛化能力和更高的预测精度。
In order to forecast the gas emission quantity of coal mine,based on the combination of the rough set and im-proved extreme learning machine,this paper absorbs the advantages of data reduction of rough set on the sample data screen-ing,and it takes full advantages of the characteristics of fast training speed and good generalization performance of extreme learning machine.It is combined with genetic algorithm theory to choose the optimal input weight matrix and the deviation of hidden layer to avoid errors caused by random.Taking advantage of the program to determine the number neurons of hidden layer is more accurate than relying on experience.In a practical application,we choose the gas content of coal layer,buried depth of coal layer,the thickness of coal layer,the spacing of mining layer,daily output of working face as the impact param-eter of the forecast.The results showed that the maximum relative error of the prediction model is 5 .687 1%,the minimum relative error is 0,the average relative error is 2.582 7%,it has stronger generalization ability and higher prediction preci-sion than prediction model of the original one.
出处
《世界科技研究与发展》
CSCD
2014年第6期638-642,共5页
World Sci-Tech R&D
关键词
粗糙集
极限学习机
瓦斯涌出量
遗传算法
数据约简
样本数据
Rough sets
Extreme learning machine
Gas emission
Genetic algorithm
Data reduction
Sample data
