摘要
为了提高径向基(RBF)网络预测瓦斯涌出量的泛化能力,提出QPSO-RBF模型。该模型采用量子粒子群(QPSO)算法优化RBF网络隐层基函数中心、扩展系数以及输出权等初始参数,将网络参数编码为QPSO学习算法中的粒子个体,在全局空间中搜索最优适应值参数。其中,RBF网络选取5-3-1的精简结构,采用5个变量作为影响因子预测瓦斯涌出量。结果表明,经QPSO优化后的RBF网络模型预测结果稳定且唯一,其泛化指标平均相对变动值(ARV)为0.012 2。与PSO-RBF、RBF模型预测结果比较,QPSO-RBF模型的泛化能力和网络训练速度优于前2种;预测精度约为PSO-RBF模型的1.5倍、RBF模型的4倍。
In order to improve the generalization ability of RBF network to predict gas emission, a QPSO- RBF model is proposed. This model uses the QPSO to optimize the initial parameters of RBF network, namely, to optimize the basis function centers of RBF hidden layer, expansion coefficient, and output weights, encode the network parameters as the particles individuals in the QPSO learning algorithm, and search for the best fitness value parameters in global space. The RBF network selects the 5 - 3 - 1 stream- lined structure, and uses five variables as the impact factors to predict the gas emission. The experiments show that the RBF network model optimized by QPSO can produce an only and stable prediction result, and that its ARV (Average Relative Variance) of generalization index is 0.012 2. Comparied with the pre- diction results of PSO-RBF, RBF model, the generalization ability and the training speed of QPSO-RBF model is better than the first two models, and the prediction accuracy about 1.5 times of PSO-RBF model, four times of RBF.
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2012年第12期29-34,共6页
China Safety Science Journal
基金
国家安全生产监督管理总局安全生产科技发展指导性计划项目(06-472)
河北省教育厅科学研究基金资助(Z2006439)