摘要
针对煤矿工作面瓦斯突出系统的非线性、复杂性、不确定性等特点,提出了狼群-蛙跳算法与Elman神经网络结合的耦合方法用于煤与瓦斯突出的非线性映射.依据预测残差方差比检验策略确定动态训练样本集,可实时的对Elman网络权值、阈值寻优,建立了基于狼群-蛙跳最优估计的瓦斯突出预测模型,并结合矿井监测数据进行实例分析.试验结果表明:采用动态的训练样本和网络结构建立的狼群-蛙跳与ENN模型跟踪能力好,收敛速度快,有效地实现了瓦斯突出危险性预测.
Aiming at the nonlinearity,complexity and uncertainty of the gas outburst in coal mine working face,this paper put forward a method that uses wolf pack algorithm and shuffled frog leaping coupling algorithm to optimize Elman neural network for performing nonlinear mapping.The dynamic set of training samples were built based on strategy for predicting variance ratio of residual errors.The Elman network weights and thresholds can be optimized in real time.The gas outburst prediction model was established based on the optimal estimation of wolf pack algorithm and shuffled frog leaping coupling algorithm,and experimental case analysis was carried out by combining with mine actual monitoring data.The results of experiment show that the wolf pack algorithm and shuffled frog leaping coupling algorithm and ENN model of coal and gas has good traceability and fast convergence rate by using the dynamic samples and network structure,which can realize the gas outburst prediction effectively.
出处
《辽宁工程技术大学学报(自然科学版)》
CAS
北大核心
2018年第3期653-659,共7页
Journal of Liaoning Technical University (Natural Science)
基金
沈阳市科技局科技重大攻关(创新专项)基金(F15-007-2-00)
关键词
瓦斯突出
预测模型
非线性
残差方差比策略
狼群-蛙跳算法
coal and gas outburst
prediction model
nonlinear
strategy for predicting variance ratio of residual errors
wolf pack algorithm and shuffled frog leaping algorithm