Abstract: In recent years，mine disasters are occuring frequently，especially gas accidents which are often reported in newspapers.Gas accidents are usually accompanied by a high concentration of the gas，so gas concentration predicting of the future moment is an effective means of gas accident forecasting and has very important significance to production safety in coal mines.This paper simplifies C-C method on chaos theory，and reconstructs phase space using time series of gas concentration values by monitoring from five coal mines.The paper chooses an appropriate value for the delay time and the embedding dimension according to the real values of gas concentration，and then forecasts the gas concentration of the next moment using weight one-rank local-region method.The experimental results show time series?of gas concentration have obviously the chaotic property.The results also show moderate calculation and better forecast results when the length of time series is 500.The mean square error of forecast results of 100 abnormal gas concentration values is 0.122 024，thus the method can be used for the prediction of gas accidents and provide basis for decision making in coal mines，for example，taking measures such as ventilationin in a timely manner.

Key words: time series of gas concentration, chaos, reconstruct phase space

摘要： 近年来各种矿难频发，特别是瓦斯事故时见于报端。瓦斯事故通常伴随着较高的瓦斯浓度，因此，预测未来时刻的瓦斯浓度是预测瓦斯事故的有效手段，对煤矿的安全生产具有十分重要的意义。对混沌理论中的C-C方法进行简化，并用这种方法对5大煤矿的瓦斯浓度监测数据构成的时间序列进行相空间重构，依据数据的实际情况确定其最佳时延和嵌入维，然后用加权一阶局域法对下一时刻的瓦斯浓度进行预测。实验结果表明瓦斯浓度时间序列具有明显的混沌特性，且当时间序列长度为500时，计算量适中且预测结果较优，对500个异常瓦斯浓度预测的均方误差达到0.122 024，从而可用于瓦斯事故的预测，为煤矿及时采取通风等措施提供决策依据。

关键词: 瓦斯浓度时间序列, 混沌, 相空间重构