论文部分内容阅读
为分析安全预测中时间序列的非平稳特性并提高预测精度,提出基于集合经验模态分解(EEMD)、相空间重构(PSR)及神经网络的预测建模方法。首先应用EEMD方法将时间序列分解成若干具有不同周期性或趋势性的分量,通过C-C方法计算各分量的最佳嵌入维数和延迟时间;然后分别进行相空间重构;再应用Elman神经网络对各分量进行训练并建立预测模型;最后将各分量预测结果叠加得到最终预测值。用该方法分析反映煤矿安全生产的关键性指标——煤炭生产百万吨死亡率。结果得到具有长期趋势性和周期性波动的5个分量,预测相对误差为-0.11%~0.20%;外推预测表明,中国煤炭生产百万吨死亡率将保持持续下降趋势,至2020年将下降到0.05以下。
In order to analyze the non-stationary characteristics of time series and improve the prediction accuracy, a predictive modeling method based on collective empirical mode decomposition (EEMD), phase space reconstruction (PSR) and neural network is proposed. Firstly, the EEMD method is used to decompose the time series into several components with different periodicity or trend. The best embedding dimension and delay time of each component are calculated by CC method. Then phase space reconstruction is carried out respectively. Then Elman neural network The components are trained and the prediction model is established. Finally, the predicted results of the components are summed to obtain the final prediction value. This method is used to analyze the key indicator reflecting the coal mine safety production - the mortality rate of 1 million tons of coal produced. The results show that there are 5 components with long-term trend and cyclical fluctuation, and the relative error of prediction is -0.11% -0.20%. The extrapolation prediction shows that the mortality rate of 1 million tons of coal in China will keep decreasing and will decrease by 2020 To 0.05 the following.