基于LS-SVR的混沌时间序列多参数自适应预测

故障预测对保障复杂设备的安全可靠工作具有重要意义,但往往难以建立起准确的解析形式的数学模型,因此常常依赖于通过观测所获得的混沌时间序列进行预测分析.为了提高预测的有效性和准确性,基于支持向量机预测理论,提出考虑全部相关多参数混沌时间序列中的信息,进行多参数相空间重构产生训练样本,并建立了多参数自适应最小二乘支持向量回归预测模型.以某设备三个相关参数的仿真混沌时间序列为例进行了预测实验,结果表明该方法有较好的预测精度,是一种有效的预测方法.

Multi-parameter Adaptive Prediction for Chaotic Time Series Based on Least Squares Support Vector Regression

Fault prediction is of great importance to ensure complex equipments＇ safety and reliability, but it is difficult to establish mathematical model of complex equipment. So complex equipments are usually predicted and analyzed by their observed chaotic time series. In order to improve the prediction availability and veracity, based on complex equipments inherent characteristics of chaotic and support vector machine prediction theory, it is suggested that all the information from relative parameters＂ time series be considered, the training samples be obtained through phase space reconstruction of chaotic time series of multi-parameter and the multi-parameter adaptive prediction model based on least squares support vector regression be established. Prediction experiments are made by taking simulation chaotic time series of three parameters of a certain complex equipment as an example. The results indicate that the method offers an effective prediction with its better precision.