一种EKF-WLS-SVR与混沌时间序列分析的瓦斯动态预测新方法

发布时间：2018-06-15 14:08

本文选题：动态预测 + 瓦斯浓度　；参考：《传感技术学报》2015年01期

【摘要】：针对瓦斯浓度时间序列高度的混沌特性,采用微熵率法同步确定最优的嵌入维数与延迟时间,还原瓦斯涌出系统状态空间。以无线传感网络系统采集并经降噪处理后的瓦斯浓度序列作为样本。提出利用带有整定因子的扩展卡尔曼滤波器(EKF)对加权最小二乘支持向量回归机(WLS-SVR)的正则化参数γ与核参数σ进行快速寻优,并依据周期性更新的训练样本建立基于EKF-WLS-SVR耦合算法的动态预测模型以精确预测后续时间点的瓦斯浓度。通过MATLAB进行仿真,结果表明:EKF滤波器对提高WLS-SVR的拟合精度与学习效率方面有很大的帮助。相比于其他模型,该耦合模型具备更高的预测精度与更强的鲁棒特性,有较高的实用价值。
[Abstract]:According to the chaotic characteristics of the gas concentration time series, the optimal embedding dimension and delay time are determined synchronously by the micro-entropy rate method, and the state space of the gas emission system is reduced. The gas concentration sequence collected by wireless sensor network system and treated by noise reduction is taken as the sample. An extended Kalman filter with tuning factor (EKF) is proposed to rapidly optimize the regularization parameter 纬 and kernel parameter 蟽 of weighted least squares support vector regression machine (WLS-SVR). A dynamic prediction model based on EKF-WLS-SVR coupling algorithm is established based on periodically updated training samples to accurately predict the gas concentration at subsequent time points. The simulation results with MATLAB show that the WLS-SVR filter has great help in improving the fitting accuracy and learning efficiency of WLS-SVR. Compared with other models, the coupling model has higher prediction accuracy and stronger robustness, and has higher practical value.
【作者单位】：辽宁工程技术大学电气与控制工程学院;国网辽宁省电力有限公司锦州供电公司;
【基金】：国家自然科学基金项目(51274118) 辽宁省教育厅基金项目(L2012119) 辽宁省科技攻关项目(2011229011)
【分类号】：TD712

【相似文献】