基于W-RBF的瓦斯时间序列预测方法  

Prediction of gas emission time series based on W-RBF

作　　者：高莉[1] 胡延军[2] 于洪珍[2] 

机构地区：[1]徐州师范大学电气工程及自动化学院,江苏徐州221116 [2]中国矿业大学信息与电气工程学院,江苏徐州221008

出　　处：《煤炭学报》2008年第1期67-70,共4页Journal of China Coal Society

基　　金：江苏省自然科学基金资助项目(BK2001073);徐州师范大学重点资助项目(06XLA25)

摘　　要：对煤矿监测监控系统采集到的瓦斯历史数据进行特征分析,通过小数据量法判定瓦斯时间序列为混沌时间序列.利用混沌时间序列的特性确定RBF神经网络的输入节点个数.提出了基于W-RBF的瓦斯时间序列预测方法.该方法将小波的多分辨率特性与RBF神经网络相结合,以提高预测精度.仿真结果表明,该方法不仅能预测出瓦斯时间序列的变化趋势,还可以保证预测值的精度,预测值与真实值对比,绝对误差最大为0.1%,且92个采样点的预测值与真实值相一致.Gas forepassed data collected from coal mine monitoring system were analyzed for their character. Then, small-data-sets were used to determine gas time series was chaos time series. Furthermore, the input neuron number of RBF neural network was determined by the character of chaos time series. A novel method of gas time series prediction method based on W - RBF was proposed on those processes, this method combine the multi-differentia- tion character of wavelet with RBF neural network. The simulation results indicate that the prediction results of W - RBF is accurate. The maximal absolute error of W - RBF is 0. 1%, and 92 sampling gain 0 error

关 键 词：瓦斯 W-RBF 时间序列预测 混沌 

分 类 号：TD712.5[矿业工程—矿井通风与安全]