基于卡尔曼滤波算法的瓦斯浓度时间序列预测分析

Gas concentration time series prediction analysis based on Kalman filter algorithm

针对自回归AR(p)模型在进行非平稳瓦斯浓度时间序列预测分析时存在精确度不高的问题,文章采用卡尔曼滤波算法动态地估算出模型参数值,在推算过程中将模型参数作为状态向量。实例分析表明基于卡尔曼滤波算法的AR(p)模型优于单一的AR(p)模型,大幅度地提高了模型的预测精度,预测效果远好于BP神经网络、支持向量机和ARMA等模型,值得借鉴。

autoregressive AR（p） model accuracy is not high in prediction analysis of non -stationa-ry time series of gas concentration .The article adopts the Kalman filter algorithm to dynamically estimate the model parameter values ,in the process of calculation ,the model parameters as the state vector .Exam-ple analysis shows that AR（p） model based on the Kalman filter algorithm is better than the single AR （p） model, greatly improve the prediction accuracy of the model ,to predict the effect is much bet-ter than BP neural networks , support vector machines and ARMA models , etc., it is worth learning from.