摘要
为提高煤与瓦斯突出强度的预测精度,提出了一种基于核主成分分析和联合改进灰狼算法优化量子门线路神经网络的KPCA-IGWO-QGCNN煤与瓦斯突出强度预测方法,确定煤与瓦斯突出强度的主要影响因素,采用核主成分分析降低瓦斯突出强度影响指标的维数,简化神经网络结构;量子门线路神经网络具有量子并行计算的优势,可有效提高信息处理的运算速度并扩大信息的存储容量,具有较好的收敛能力与鲁棒性;通过非线性控制参数与粒子群思想联合改进的灰狼算法提高模型全局寻优能力,优化量子门线路神经网络的网络参数。通过对比分析该模型与BP神经网络、QGCNN、GWO-QGCNN神经网络模型的预测结果,说明该模型泛化能力强,预测精度高。
In order to improve the prediction accuracy of the coal mine gas outburst strength, this paper proposes a joint improved grey wolf algorithm based on kernel principal component analysis and neural network to optimize the quantum gate circuit(KPCA-IGWO-QGCNN) of coal mine gas outburst intensity forecast method, to determine the main factors influencing coal mine gas outburst strength. Kernel principal component analysis is used to reduce the dimension of the gas outburst strength index and simplify the structure of neural network. Quantum gate circuit neural network has the advantage of quantum parallel computing, which can effectively improve the speed of information processing and expand the storage capacity of information, and has good convergence ability and robustness. The global optimization ability of the model is improved by combining the nonlinear control parameters with particle swarm optimization improve the grey wolf algorithm, and the network parameters of the quantum gate ciucuit neural network are optimized. By comparing the prediction results of this model with BP, QGCNN and GWO-QGCNN, it is shown that this model has strong generalization ability and high prediction accuracy.
作者
付华
孟庭儒
阎馨
卢万杰
FU Hu;MENG Ting-ru;YAN Xin;LU Wan-jie(Faculty of Electrical and Control Engineering,Liaoning Technical University,Fuxin 123000,China;School of Mechanical Engineering,Liaoning Technical University,Fuxin 123000,China)
出处
《控制工程》
CSCD
北大核心
2021年第9期1731-1737,共7页
Control Engineering of China
基金
国家自然科学基金资助项目(71771111,51974151,61601212)
辽宁省教育厅项目(LJ201QL012)
辽宁省自然科学基金指导计划项目(20180550438)。
作者简介
付华(1962-),女,辽宁阜新人,博士,教授,主要从事现代传感技术及系统、煤矿瓦斯智能检测和控制工程等方面的教学与科研工作;通信作者:孟庭儒(1996-),女,辽宁铁岭人,研究生,主要研究方向为机器学习、信息处理与模式识别等(Email:18242989703@163.com)。
