摘要： 针对煤矿瓦斯涌出量的多影响因素预测问题,引入荧光因子以自适应调整搜索步长,用于改善基本萤火虫算法后期收敛速度慢及容易陷入局部最优的缺陷。将改进后的自适应步长萤火虫算法与Elman动态反馈神经网络相结合,用于辨识瓦斯涌出非线性系统。通过实时对网络的权值、阈值进行全局寻优,建立基于ASGSO-ENN耦合算法的绝对瓦斯涌出量预测模型。利用矿井监测到的各项历史数据进行实验,结果表明,该模型的预测均方根误差为0.103 4,平均相对变动值为0.000 387。相比于其他工程常用的预测模型,具备更高的预测精度与更强的泛化能力。

关键词: 绝对瓦斯涌出量, 非线性系统, 预测模型, 自适应步长萤火虫群优化, Elman神经网络, 动态反馈

Abstract: For the multifactor gas emission quantity predication problem of coal mine,the basic Glowworm Swarm Optimization(GSO)algorithm suffers slow convergence rate in the late stage and is prone to be stuck in local optimum.The luciferin-factor is introduced in this paper to achieve self-adaption of the search step length.The proposed self-Adaptive Step Glowworm Swarm Optimization(ASGSO)algorithm is combined with dynamic feedback Elman Neural Network(ENN) to perform identification of the non-linear gas emission system.By globally searching the optimum of the network weights and thresholds in real-time,the ASGSO-ENN-based model is established by the coupling algorithm to predict the absolute gas emission quantity.The experiments are carried out with the historical monitoring data of the mine and prediction results show that,the root mean square error is 0.103 4 and the average relative variance is 0.000 387.The proposed model is practically useful and outperforms other commonly used models in terms of prediction accuracy and generalization ability.

Key words: absolute gas emission quantity, non-linear system, prediction model, Self-adaptive Step Glowworm Swarm Optimization(ASGSO), Elman Neural Network(ENN), dynamic feedbac

中图分类号: 

• TP301.6