Dynamic prediction of gas concentration based on time series

Existing gas concentration prediction methods could only achieve static gas concentration prediction, could not update with accumulation of gas data, as a result, the prediction results were not timeliness. In view of the above problem, a dynamic prediction method of gas concentration based on time series was proposed. Firstly, the method uses multi-resolution characteristic of wavelet decomposition technique to decompose the gas concentration time series to different scales to make the time series smooth. Then it adopts auto regressive and moving average(ARMA) model constructed by real-time and dynamic data to predict mine gas concentration in the future time by use of gas concentration change trend in the past time, so as to obtain time series prediction results. Finally, in order to improve the accuracy of the gas concentration prediction, the prediction results of the ARMA model and mine environment parameters are input into the trained BP neural network model, and the prediction results are corrected by the BP neural network model, so as to obtain final gas concentration prediction value. The test results show that the method can accurately predict the mine gas concentration, and the average relative error of gas concentration prediction is reduced from 8% to 5%.