Hybrid Adaptative System of Gas Concentration Prediction in Hard-Coal Mines

Hybrid and adaptative system of gas concentration prediction in hard-coal mines is presented in the paper. The system widens functionality of the SMP-NT system which monitors gas concentration in mining excavations based on data collected from sensors. The SMP-NT system has also ability to automatic cut off electric energy in the case of explosion risk identification. A task of the prediction system described in the paper is to anticipate hazards due time in advance. A first part of the prediction system uses linear and nonlinear prediction methods. Moreover it makes use of a solution of metalearning type which negotiates the system final forecast between linear and nonlinear methods. A second part of the system uses knowledge base of typical situations that lead to explosion or fire risk increase. Thanks to Dynamic Time Warping methodology, continuous monitoring of methane and other gases concentration time series is conducted from the current time series similarity to dangerous situations point of view.