基于支持向量机(SVM)的回采工作面瓦斯涌出混沌预测方法研究

Research on Chaotic Forecasting Method for Gas Emission from Working Face Based on SVM Theory

针对瓦斯涌出传统的线性预测方法存在的问题,根据瓦斯涌出时间序列固有的确定性和非线性,利用混沌动力系统的相空间延迟坐标重构理论,结合基于机器学习理论的支持向量机(SVM),建立基于SVM理论的瓦斯涌出混沌时间序列预测模型。经Ⅱ1024回采工作面瓦斯涌出时间序列仿真计算,仿真结果显示该预测模型具有比传统的回归方法更好的泛化能力,预测方法具有很高的预测精度。同时,该模型具有以往传统机器学习的瓦斯涌出预测模型建立简便、训练速度快等优点。由于充分考虑瓦斯涌出时间序列的混沌性,并利用SVM预测的优良特性,使得预测更科学。

Aimed at the deficiencies in traditional linear prediction method for gas emission, a chaotic time series prediction model for gas emission based on support vector machine（SVM） theory is built by using the phase space delay coordinates reconstruction theory of chaotic dynamical systems with consideration on the inherent determinacy and nonlinear nature of the time series of gas emission. After a simulation calculation to the time series of gas emission in Ⅱ 1024 working face, the results show that the prediction model possesses higher generalization ability than the traditional regression methods, and has high prediction accuracy. Moreover, the model has the merits of easily being built and high training-speed, etc.. Due to a full consideration on the chaos of time series of gas emission and a good utilization of the excellent prediction features of SVM, the prediction turns to be more scientific.