The Application Of Wavelet Analysis In Gas Emission Prediction

The coal mine gas disaster is one of the most dangerous five coal mine natural disasterswhich are kinds of serious threats to the safety of coal production. The monitoring andprediction of Gas Emission which is an important indicator of the gas disaster are particularlyimportant to prevent gas disaster accidents. From the point of view of the whole system, theGas Emission is a complex nonlinear dynamical system, as a kind of Time series, its size andorder contains a large number of dynamic evolution traces and characteristics of the system.In this paper, Mine Gas Emission was predicted based on combined multi-resolution analysisof wavelet theory and wavelet packet theory of the wavelet theory. The main contents andconclusions are as follow:The Gas Emission which is a non-stationary Time Series was decomposed into severallayers of approximate sense stationary Time Series by the analysis of the waveletmulti-resolution. A prediction model(multi-resolution analysis of the AR model) wasestablished using the stationary time series AR model for the single reconstruct sequence, andthe impact of the prediction was analyzed in different basis functions or same basis functionsbut different numbers of decomposition levels. From simulation, predictions based onmulti-resolution analysis were better than the basis function prediction.Multi-resolution analysis only can decompose timeline in a fixed frequency space,information reflected in some of the higher spatial frequency of Gas Emission System Statuscharacterized could has been lost in analysis of high time resolution of the time sequence,because of selected relatively low frequency scale. So that, an Adaptive chosen-band waveletpacket transform was selected in this paper for the prediction of chaotic time series of GasEmission. The traditional Wavelet Packet-chaotic model was improved based ondiscrimination of Gas Emission Chaos Characteristics. The weighing was introduced inprediction results of each sequences reconstructed by wavelet packets to establish theweighted wavelet packet-chaotic prediction model, which could not only improve theprediction accuracy also improve the instability of the prediction error, and increase the predictable range. Weighted best wavelet packet-chaotic prediction model decomposed byBest Wavelet Packet was proposed in this paper, which reserved the advantages of weightedwavelet packet-chaotic prediction model and reduces the computational amount.Finally, the weighted one-rank local law in Best Wavelet Packet-chaotic predictionmodel was replaced by Wavelet Neural Network, the Best wavelet packet-Chaos-waveletneural network model was established by using the powerful nonlinear mapping ability ofWavelet Neural Network, to predict Gas Emission. The result of the simulation experimentsand comparative analysis illustrated: the Weighted best wavelet packet-chaotic predictionmodel and Best wavelet packet-Chaos-wavelet neural network model showed a higher andsatisfied prediction accuracy, could be promoted and practiced.