摘要
基于灰熵理论和RBF神经网络理论,提出了一种改进的灰色神经网络深部煤层瓦斯含量预测模型。该模型首先利用灰熵关联度确定影响深部煤层瓦斯含量的主控因素,构建多个GM预测模型进行精度分析,寻求最优的灰色预测模块对分析系统进行一次预测,再利用灰色模型白化微分方程解序列相邻两元素分别与相应期望值作差,构建一个差值序列作为RBF神经网络输出对分析系统进行二次预测,得到的差序列预测结果的差值即为深部煤层瓦斯含量的预测值,从而构建了基于差值GM-RBF神经网络组合模型的深部煤层瓦斯含量预测体系。实际应用表明:差值GM-RBF神经网络组合模型的精度评价指标MAE、MAPE、RMSE、RRMSE分别为0.233 1、3.25%、0.2778、4.04%,远优于单一灰色、RBF模型;与传统GM-RBF组合模型相比,MAE和MAPE分别减小了23.8%和22.1%,RMSE和RRMSE分别减小了20.5%和17%。由此可见,以差值结合法将最优灰色模块与RBF神经网络有效结合起来的瓦斯含量预测体系增强了模型的泛化能力和数据利用率,精度更高,稳定性更好,能够满足深部煤层瓦斯含量准确预测的要求,为深部煤与瓦斯安全高效开采提供依据。
The paper is inclined to present an improved grey model( GM) with the artificial neural network to predict the gas content in the deep coal seam by resorting to the grey entropy theory and the artificial neural network with the radial basis function( RBF). In the said grey model( GM),the primary factor of the gas content in the deep coal seam can be determined by using the correlation of the grey entropy. The predictions of some other grey models have also been built up to perform the precision analysis.What is more,an optimized grey module for prediction has been established to perform the first prediction of the analysis system.And,then,it would be possible to set up a difference sequence as the output of the RBF neural network model by the two adjacent elements picked up from the albinism differential equation results of GM model minus the corresponding expected value. The said method can also be adopted to do the second prediction for the system. The numerical simulation results we have gained suggest that the evaluation parameters of MAE,MAPE,RMSE and RRMSE in the above mentioned GM-RBF model can respectively reach 0. 233 1,3. 25%,0. 277 8 and 4. 04%,which can be said far better in comparison with the single prediction model and RBF model. Besides,whereas the MAE and MAPE tend to drop to23. 8% and 22. 1%,respectively,the RMSE and RRMSE should be decreased to 20. 5% and 17%,correspondingly. The results thus gained imply that the combination of the optimized grey module and the artificial neural network with RBF are in a position to enhance the generalization and data utilization for the model. Moreover,the improved GM-RBF model can be said to enjoy a higher accuracy and stability so as to be able to meet the demand for predicting the gas content in the deep coal seam. In addition,the aforementioned results suggest that the improved model can be expected to provide a useful tool for the coal mining in the deep coal seam.
出处
《安全与环境学报》
CAS
CSCD
北大核心
2017年第6期2050-2055,共6页
Journal of Safety and Environment
关键词
安全工程
深部煤层
瓦斯含量预测
灰熵关联度
RBF神经网络
差序列
safety engineering
deep coal seam
gas contentprediction
grey entropy correlation grade
RBFneur/d network
difference sequence